A model is still running

A model is still running or downloading. Leaving will stop the current operation. Stay on this page to let it finish.

Skip to content
LocalAI.toolsBeta

Join the waitlist

Get an email when the new apps and model runs launch.

No account needed.

Stable Audio 3.0

Generate short stereo sound effects from English descriptions locally on your desktop.

Desktop inferencePrivate · Runs on your computer
About this model

Generate short stereo sound effects from English descriptions locally on your desktop.

Created by
Stability AI
Model released
May 20, 2026
Source repository created
May 17, 2026
Model weights · FP32
3.5 GB

Best for

  • Sound effects for creative drafts, games and videos
  • Local text-to-audio experimentation

Keep in mind

  • Generated sounds may omit, merge or misinterpret requested events; review before use.
  • English prompts and short effects are tested; music, speech, editing and long-form generation are not exposed.
  • No content safety classifier is included; generated content may be unsuitable or biased.
  • Long or ambiguous descriptions can lose details. This adapter limits prompts to 2000 characters and audio to 1–30 seconds.
  • Requires accepting the upstream gated-model terms and authorized Hugging Face credentials for download. Gemma terms apply to the text encoder.
  • Only current Apple Silicon Mac CPU/Metal execution is measured; other platforms remain unverified.
  • FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark.

Processed on your deviceYour input is processed locally in the desktop app. Model and runtime files download from external hosts.

Target platforms · verification required

  • macos · aarch64, x86_64
  • windows · x86_64

Recommended configuration

FP32 · mps (verified platform only)

Sources & technical details
Original model
https://huggingface.co/stabilityai/stable-audio-3-small-sfx
Release verification
Verified
Release evidence
Official announcement explicitly includes Small SFX release; repository creation is separate.
Source updated
Oct 6, 2026
Native model files
https://huggingface.co/stabilityai/stable-audio-3-small-sfx
Files updated
Oct 6, 2026
Artifact revision
05ca3857881c94a2b84026b1111c537d78f5f9f0

Desktop app

Sample result

Stable Audio 3.0 Small SFX

Run Stable Audio 3.0 on your computer

Download the desktop app to load this model, provide input and view results locally.

Public desktop downloads are coming soon.

Device requirements

macOS Apple SiliconCPU / FP32Tested16.0 GiBRAM estimate
macOS Apple SiliconMPS / FP32Tested16.0 GiBRAM estimate
Windows Intel / AMDCPU / FP32Unverified16.0 GiBRAM estimate
macOS Intel / AMDCPU / FP32Unverified16.0 GiBRAM estimate
If your device falls short

MemoryAdvisory only: close other apps, reduce image/tile/token/batch settings or choose a smaller verified model/precision. CPU can need more RAM than GPU.

GPUIf the selected GPU runtime fails, existing CPU fallback may recover when enabled, but CPU has its own memory/runtime needs.

PlatformUse a compatible OS/architecture/runtime or another model. These estimates do not add hardware detection or block runs.

Benchmark evidence

Saved desktop measurements apply to the recorded hardware and runtime. Fallback cases report the device that actually ran; they do not prove GPU support on the requested device.

Stable Audio 3.0 Small SFX

16 configurations tested
Passed16
Failed0
TestedNov 4, 2026
Runtimestable-audio-3 0.1.0+790793b

Quality criteria

No explicit quality gate recorded. Review per-case evidence; full certification is pending.

PrecisionDeviceResultPlatformInference / loadTask metric
FP32mpsAcceleratedPassedFastmacos · aarch64729 ms / 0 ms0.146 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64731 ms / 0 ms0.146 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64734 ms / 0 ms0.147 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64738 ms / 0 ms0.148 RTF
FP32mpsAcceleratedPassedFastmacos · aarch64745 ms / 0 ms0.149 RTF
FP32mpsAcceleratedPassedFastmacos · aarch641219 ms / 3583 ms0.244 RTF
FP32mpsAcceleratedPassedFastmacos · aarch641255 ms / 3425 ms0.251 RTF
FP32mpsAcceleratedPassedFastmacos · aarch641271 ms / 3419 ms0.254 RTF
FP32cpuPassedFastmacos · aarch642025 ms / 3033 ms0.405 RTF
FP32cpuPassedFastmacos · aarch642044 ms / 2970 ms0.409 RTF
FP32cpuPassedFastmacos · aarch642082 ms / 0 ms0.416 RTF
FP32cpuPassedFastmacos · aarch642087 ms / 3203 ms0.417 RTF
FP32cpuPassedFastmacos · aarch642095 ms / 0 ms0.419 RTF
FP32cpuPassedFastmacos · aarch642095 ms / 0 ms0.419 RTF
FP32cpuPassedFastmacos · aarch642216 ms / 0 ms0.443 RTF
FP32cpuPassedFastmacos · aarch642241 ms / 0 ms0.448 RTF
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-04T18:09:51.095513+00:00
Environment
{ "os": "macos", "arch": "aarch64", "hardware": "Apple M3 Max", "osVersion": "27.2", "runtime": "Stable Audio 3 pinned 790793b / Torch 2.7.1 / Transformers 5.8.0", "interpreter": "App-owned CPython 3.13.18 private probe venv" }
Runtime
stable-audio-3 0.1.0+790793b
Artifact revision
05ca3857881c94a2b84026b1111c537d78f5f9f0
Fixture
{
  "file": "model-samples/desktop/stable-audio3/dog.txt",
  "size": 44,
  "sha1": "6b565150c3d1214cba75722f7222b15b3936efae",
  "sha256": "340c80af24ba2ee557fb3e0dc5da9e48deca7bf1779ee3ad1a518b437d2f001f",
  "url": "https://assets.localai.tools/samples/v1/files/340c80af24ba2ee557fb3e0dc5da9e48deca7bf1779ee3ad1a518b437d2f001f/dog.txt"
}
Rerun when
Checkpoint/runtime/adapter/config changes; Fixture or quality gate changes; Relevant device/backend behavior
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a dog barking.",
          "score": 0.999996542930603
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000013630385637952713
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000001114351562137017
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 7.826943146937992e-7
        },
        {
          "label": "A song with music and singing.",
          "score": 1.709511678882336e-7
        }
      ],
      "expected": "The sound of a dog barking.",
      "rms": 0.0676775649189949
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3203.4123339981306,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 2087.4079579953104,
      "durationMs": 2092.1912909834646,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.3953151950239104,
      "realTimeFactor": 0.41748159159906206,
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.0676775649189949,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 5295.603624981595,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.9993638396263123
        },
        {
          "label": "A song with music and singing.",
          "score": 0.0005996496765874326
        },
        {
          "label": "The sound of a dog barking.",
          "score": 0.000020735013094963506
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.000011260440260230098
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000004544238890957786
        }
      ],
      "expected": "The sound of rain falling and water splashing.",
      "rms": 0.014432086609303951
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 2215.959916007705,
      "durationMs": 2219.7543749934994,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.256358503545519,
      "realTimeFactor": 0.443191983201541,
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.014432086609303951,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 2219.7543749934994,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.9999579191207886
        },
        {
          "label": "A song with music and singing.",
          "score": 0.000031964227673597634
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.00000571110876990133
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000004076204731973121
        },
        {
          "label": "The sound of a dog barking.",
          "score": 3.8981437455731793e-7
        }
      ],
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "rms": 0.055390939116477966
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 2241.4129590033554,
      "durationMs": 2244.942459015874,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.230735741897045,
      "realTimeFactor": 0.4482825918006711,
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.055390939116477966,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 2244.942459015874,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a dog barking.",
          "score": 0.999996542930603
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000013630385637952713
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000001114351562137017
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 7.826943146937992e-7
        },
        {
          "label": "A song with music and singing.",
          "score": 1.709511678882336e-7
        }
      ],
      "expected": "The sound of a dog barking.",
      "rms": 0.0676775649189949
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 2094.874875008827,
      "durationMs": 2098.7943329964764,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.3867773964203622,
      "realTimeFactor": 0.4189749750017654,
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.0676775649189949,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 2098.7943329964764,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.9993638396263123
        },
        {
          "label": "A song with music and singing.",
          "score": 0.0005996496765874326
        },
        {
          "label": "The sound of a dog barking.",
          "score": 0.000020735013094963506
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.000011260440260230098
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000004544238890957786
        }
      ],
      "expected": "The sound of rain falling and water splashing.",
      "rms": 0.014432086609303951
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 2082.1474170079455,
      "durationMs": 2086.057584005175,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.401366953731365,
      "realTimeFactor": 0.4164294834015891,
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.014432086609303951,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 2086.057584005175,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.9999579191207886
        },
        {
          "label": "A song with music and singing.",
          "score": 0.000031964227673597634
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.00000571110876990133
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000004076204731973121
        },
        {
          "label": "The sound of a dog barking.",
          "score": 3.8981437455731793e-7
        }
      ],
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "rms": 0.055390939116477966
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 2095.262290997198,
      "durationMs": 2099.1044999973383,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.3863360790120223,
      "realTimeFactor": 0.4190524581994396,
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.055390939116477966,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 2099.1044999973383,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a dog barking.",
          "score": 0.999996542930603
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000013630385637952713
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000001114351562137017
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 7.826943146937992e-7
        },
        {
          "label": "A song with music and singing.",
          "score": 1.709511678882336e-7
        }
      ],
      "expected": "The sound of a dog barking.",
      "rms": 0.0676775649189949
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 2969.9729999992996,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 2044.3073750066105,
      "durationMs": 2048.420042003272,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.4458161532503553,
      "realTimeFactor": 0.4088614750013221,
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.0676775649189949,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 5018.393042002572,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a dog barking.",
          "score": 0.999996542930603
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000013630385637952713
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000001114351562137017
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 7.826943146937992e-7
        },
        {
          "label": "A song with music and singing.",
          "score": 1.709511678882336e-7
        }
      ],
      "expected": "The sound of a dog barking.",
      "rms": 0.0676775649189949
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3032.590374990832,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 2025.0489579921123,
      "durationMs": 2029.1257920034695,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.469076108143888,
      "realTimeFactor": 0.40500979159842243,
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.0676775649189949,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 5061.716166994302,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a dog barking.",
          "score": 0.9999892711639404
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000006887236850161571
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000019031698457183666
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.0000011380615205780487
        },
        {
          "label": "A song with music and singing.",
          "score": 6.937881948942959e-7
        }
      ],
      "expected": "The sound of a dog barking.",
      "rms": 0.057799357920885086
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3419.3780420173425,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 1270.7210409862455,
      "durationMs": 1277.001915994333,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 3.9347739108178668,
      "realTimeFactor": 0.2541442081972491,
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.057799357920885086,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": 3394961408,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 4696.379958011676,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.9973976612091064
        },
        {
          "label": "A song with music and singing.",
          "score": 0.0018542429897934198
        },
        {
          "label": "The sound of a dog barking.",
          "score": 0.00039657147135585546
        },
        {
          "label": "A person giving a speech.",
          "score": 0.00033967671333812177
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.000011799330422945786
        }
      ],
      "expected": "The sound of rain falling and water splashing.",
      "rms": 0.015266445465385914
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 738.2594579830766,
      "durationMs": 742.3173329734709,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 6.772686683432394,
      "realTimeFactor": 0.14765189159661532,
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.015266445465385914,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": 3394961408,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 742.3173329734709,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.9999754428863525
        },
        {
          "label": "A song with music and singing.",
          "score": 0.000013841670806868933
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000008418480319960508
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000015484582718272577
        },
        {
          "label": "The sound of a dog barking.",
          "score": 7.015021878942207e-7
        }
      ],
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "rms": 0.05962764099240303
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 733.7421249831095,
      "durationMs": 737.6897919748444,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 6.8143831868929405,
      "realTimeFactor": 0.1467484249966219,
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.05962764099240303,
      "clippedFraction": 0.000027210884353741498,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": 3394961408,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 737.6897919748444,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a dog barking.",
          "score": 0.9999892711639404
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000006887236850161571
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000019031698457183666
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.0000011380615205780487
        },
        {
          "label": "A song with music and singing.",
          "score": 6.937881948942959e-7
        }
      ],
      "expected": "The sound of a dog barking.",
      "rms": 0.057799357920885086
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 730.869375023758,
      "durationMs": 735.047417023452,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 6.841167752907237,
      "realTimeFactor": 0.1461738750047516,
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.057799357920885086,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": 3394961408,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 735.047417023452,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.9973976612091064
        },
        {
          "label": "A song with music and singing.",
          "score": 0.0018542429897934198
        },
        {
          "label": "The sound of a dog barking.",
          "score": 0.00039657147135585546
        },
        {
          "label": "A person giving a speech.",
          "score": 0.00033967671333812177
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.000011799330422945786
        }
      ],
      "expected": "The sound of rain falling and water splashing.",
      "rms": 0.015266445465385914
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 729.1464160080068,
      "durationMs": 733.2857500005048,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 6.857333301279088,
      "realTimeFactor": 0.14582928320160135,
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.015266445465385914,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": 3394961408,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 733.2857500005048,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.9999754428863525
        },
        {
          "label": "A song with music and singing.",
          "score": 0.000013841670806868933
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000008418480319960508
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000015484582718272577
        },
        {
          "label": "The sound of a dog barking.",
          "score": 7.015021878942207e-7
        }
      ],
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "rms": 0.05962764099240303
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 0,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 745.204708975507,
      "durationMs": 749.3434169737156,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 6.709565760626907,
      "realTimeFactor": 0.14904094179510138,
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.05962764099240303,
      "clippedFraction": 0.000027210884353741498,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": 3394961408,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 749.3434169737156,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a dog barking.",
          "score": 0.9999892711639404
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000006887236850161571
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000019031698457183666
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.0000011380615205780487
        },
        {
          "label": "A song with music and singing.",
          "score": 6.937881948942959e-7
        }
      ],
      "expected": "The sound of a dog barking.",
      "rms": 0.057799357920885086
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3425.416417012457,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 1255.4465830035042,
      "durationMs": 1261.141958006192,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 3.9826465480021493,
      "realTimeFactor": 0.25108931660070083,
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.057799357920885086,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": 3394961408,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 4686.558375018649,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "CLAP top-1 intended effect among 5 fixed labels + signal validity",
      "predictions": [
        {
          "label": "The sound of a dog barking.",
          "score": 0.9999892711639404
        },
        {
          "label": "A person giving a speech.",
          "score": 0.000006887236850161571
        },
        {
          "label": "The sound of rain falling and water splashing.",
          "score": 0.0000019031698457183666
        },
        {
          "label": "The sound of a wooden door creaking and slamming shut.",
          "score": 0.0000011380615205780487
        },
        {
          "label": "A song with music and singing.",
          "score": 6.937881948942959e-7
        }
      ],
      "expected": "The sound of a dog barking.",
      "rms": 0.057799357920885086
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3582.9105830052868,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inferenceMs": 1219.3501660076436,
      "durationMs": 1224.770166008966,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 4.100544814268437,
      "realTimeFactor": 0.24387003320152872,
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.057799357920885086,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "driverAllocatedMemoryBytes": 3394961408,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "05ca3857881c94a2b84026b1111c537d78f5f9f0",
      "fallback": false,
      "grade": "fast",
      "totalMs": 4807.680749014253,
      "precisionDetails": "FP32 selects the audio diffusion model and decoder; the pinned upstream T5Gemma text encoder retains BF16. Exact mixed parameter placement is recorded in every benchmark."
    }
  }
]

Stable Audio 3.0 Medium

22 configurations tested
Passed22
Failed0
TestedNov 7, 2026
Runtimestable-audio-3 0.1.0+790793b

Quality criteria

No explicit quality gate recorded. Review per-case evidence; full certification is pending.

PrecisionDeviceResultPlatformInference / loadTask metric
FP32mpsAcceleratedPassedFastmacos · aarch642182 ms / 16325 ms0.436 RTF
FP32mpsAcceleratedPassedFastmacos · aarch642248 ms / 8854 ms0.450 RTF
FP32mpsAcceleratedPassedFastmacos · aarch642262 ms / 16325 ms0.452 RTF
FP32mpsAcceleratedPassedFastmacos · aarch642276 ms / 11059 ms0.455 RTF
FP32mpsAcceleratedPassedFastmacos · aarch642307 ms / 11059 ms0.461 RTF
FP32mpsAcceleratedPassedFastmacos · aarch642400 ms / 8854 ms0.480 RTF
FP32mpsAcceleratedPassedFastmacos · aarch642454 ms / 16368 ms0.491 RTF
FP32mpsAcceleratedPassedFastmacos · aarch642648 ms / 16368 ms0.530 RTF
FP32mpsAcceleratedPassedFastmacos · aarch642913 ms / 8854 ms0.583 RTF
FP32mpsAcceleratedPassedFastmacos · aarch643039 ms / 11059 ms0.608 RTF
FP32mpsAcceleratedPassedFastmacos · aarch644530 ms / 16325 ms0.906 RTF
FP32cpuPassedFastmacos · aarch649583 ms / 7663 ms1.917 RTF
FP32cpuPassedFastmacos · aarch649617 ms / 7663 ms1.923 RTF
FP32cpuPassedFastmacos · aarch649669 ms / 7930 ms1.934 RTF
FP32cpuPassedFastmacos · aarch649676 ms / 7698 ms1.935 RTF
FP32cpuPassedFastmacos · aarch649684 ms / 7698 ms1.937 RTF
FP32cpuPassedFastmacos · aarch649744 ms / 7930 ms1.949 RTF
FP32cpuPassedFastmacos · aarch649958 ms / 7663 ms1.992 RTF
FP32cpuPassedFastmacos · aarch6410181 ms / 7698 ms2.036 RTF
FP32cpuPassedFastmacos · aarch6410534 ms / 7930 ms2.107 RTF
FP32cpuPassedSlowmacos · aarch6415349 ms / 7732 ms3.070 RTF
FP32cpuPassedSlowmacos · aarch6415935 ms / 7732 ms3.187 RTF
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-07T14:50:59.989331+00:00
Environment
{ "os": "macos", "arch": "aarch64", "hardware": "Apple M3 Max", "osVersion": "27.2", "runtime": "Stable Audio 3 pinned 790793b / Torch 2.7.1 / Transformers 5.8.0", "interpreter": "App-owned CPython 3.13.18 private probe venv", "concurrentActivity": "Pinned candidate source downloads and tracked background archival service active. CLAP scoring runs between generation cases." }
Runtime
stable-audio-3 0.1.0+790793b
Artifact revision
d4d3579395a8ac0df4ab13b5b1d7138ce7290f81
Fixture
{
  "file": "model-samples/desktop/stable-audio3/medium/medium-piano.txt",
  "size": 83,
  "sha1": "bdf829476f6d8979181472b69c8ff1e40d822a37",
  "sha256": "598e2c9b870380f0e7632de52d63bcaf3ec5343badf49ba10b38b507d057d783",
  "url": "https://assets.localai.tools/samples/v1/files/598e2c9b870380f0e7632de52d63bcaf3ec5343badf49ba10b38b507d057d783/medium-piano.txt"
}
Rerun when
Checkpoint/runtime/adapter/config changes; Fixture or quality gate changes; Relevant device/backend behavior
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.998745322227478
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0005576094845309854
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00037735013756901026
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00031569250859320164
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000004077629910170799
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.998745322227478,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 4529.804499994498,
      "durationMs": 4537.442792003276,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 1.1038003958020868,
      "realTimeFactor": 0.9059608999988995,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.06136477738618851,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11491672064,
      "loadMs": 16325.327582992031,
      "totalMs": 20862.770374995307,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9998389482498169
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00006817697430960834
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00006152190326247364
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00001817176234908402
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.000013212109479354694
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9998389482498169,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 2262.2142920008628,
      "durationMs": 2268.2996669900604,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.2102238579607087,
      "realTimeFactor": 0.45244285840017257,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.17616049945354462,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11500060672,
      "loadMs": 16325.327582992031,
      "totalMs": 18593.62724998209,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9999080896377563
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00005459109161165543
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00003499094964354299
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000013119408777129138
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0000010166077117901295
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9999080896377563,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 2182.4558329972206,
      "durationMs": 2186.170500004664,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.2909971072053157,
      "realTimeFactor": 0.4364911665994441,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.09808008372783661,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11500060672,
      "loadMs": 16325.327582992031,
      "totalMs": 18511.498082996695,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.998745322227478
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0005576094845309854
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00037735013756901026
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00031569250859320164
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000004077629910170799
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.998745322227478,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 3038.9532499975758,
      "durationMs": 3045.2377920009894,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 1.645303362269225,
      "realTimeFactor": 0.6077906499995152,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.06136477738618851,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11491672064,
      "loadMs": 11058.775999990758,
      "totalMs": 14104.013791991747,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9998389482498169
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00006817697430960834
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00006152190326247364
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00001817176234908402
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.000013212109479354694
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9998389482498169,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 2275.9341670025606,
      "durationMs": 2280.821749998722,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.19690010040364,
      "realTimeFactor": 0.45518683340051214,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.17616049945354462,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11500060672,
      "loadMs": 11058.775999990758,
      "totalMs": 13339.59774998948,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9999080896377563
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00005459109161165543
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00003499094964354299
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000013119408777129138
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0000010166077117901295
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9999080896377563,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 2306.6237500024727,
      "durationMs": 2310.336542010191,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.1676703883737605,
      "realTimeFactor": 0.46132475000049455,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.09808008372783661,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11500060672,
      "loadMs": 11058.775999990758,
      "totalMs": 13369.112542000948,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.998745322227478
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0005576094845309854
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00037735013756901026
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00031569250859320164
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000004077629910170799
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.998745322227478,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 2912.5423340010457,
      "durationMs": 2918.0512089951662,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 1.716713244518356,
      "realTimeFactor": 0.5825084668002092,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.06136477738618851,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11491672064,
      "loadMs": 8854.373374997522,
      "totalMs": 11772.424583992688,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9998389482498169
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00006817697430960834
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00006152190326247364
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00001817176234908402
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.000013212109479354694
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9998389482498169,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 2248.0111249897163,
      "durationMs": 2251.951416998054,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.224188281106871,
      "realTimeFactor": 0.4496022249979433,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.17616049945354462,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11500060672,
      "loadMs": 8854.373374997522,
      "totalMs": 11106.324791995576,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9999080896377563
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00005459109161165543
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00003499094964354299
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000013119408777129138
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0000010166077117901295
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9999080896377563,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 2399.7663340123836,
      "durationMs": 2403.8826250034617,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.08353618814214,
      "realTimeFactor": 0.47995326680247674,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.09808008372783661,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11500060672,
      "loadMs": 8854.373374997522,
      "totalMs": 11258.256000000983,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9992197751998901
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0005686170770786703
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00016254348156508058
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00004823402196052484
          },
          {
            "label": "The sound of a dog barking.",
            "score": 7.209198997770727e-7
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9992197751998901,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 10533.7145000085,
      "durationMs": 10538.495250002597,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.47466636768976084,
      "realTimeFactor": 2.1067429000017,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06947697699069977,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7930.256499996176,
      "totalMs": 18468.751749998773,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9997052550315857
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00020508802845142782
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00004371672184788622
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00003144167203572579
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000014381625987880398
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9997052550315857,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 9743.796791997738,
      "durationMs": 9746.963874989888,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.5131469905146561,
      "realTimeFactor": 1.9487593583995477,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.16417637467384338,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7930.256499996176,
      "totalMs": 17677.220374986064,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9998252987861633
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00012907838390674442
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00004292851735954173
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000018302520174984238
          },
          {
            "label": "A person giving a speech.",
            "score": 9.836167009780183e-7
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9998252987861633,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 9669.028624994098,
      "durationMs": 9672.590542002581,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.5171150271574516,
      "realTimeFactor": 1.9338057249988196,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.11127196252346039,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7930.256499996176,
      "totalMs": 17602.847041998757,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9992197751998901
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0005686170770786703
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00016254348156508058
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00004823402196052484
          },
          {
            "label": "The sound of a dog barking.",
            "score": 7.209198997770727e-7
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9992197751998901,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 10180.982165999012,
      "durationMs": 10185.204333000001,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.4911117531173254,
      "realTimeFactor": 2.0361964331998026,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06947697699069977,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7698.144041001797,
      "totalMs": 17883.3483740018,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9997052550315857
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00020508802845142782
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00004371672184788622
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00003144167203572579
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000014381625987880398
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9997052550315857,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 9676.083583006402,
      "durationMs": 9679.129916999955,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.5167379918856052,
      "realTimeFactor": 1.9352167166012806,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.16417637467384338,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7698.144041001797,
      "totalMs": 17377.27395800175,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9998252987861633
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00012907838390674442
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00004292851735954173
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000018302520174984238
          },
          {
            "label": "A person giving a speech.",
            "score": 9.836167009780183e-7
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9998252987861633,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 9683.918665992678,
      "durationMs": 9687.11204099236,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.5163199085468011,
      "realTimeFactor": 1.9367837331985356,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.11127196252346039,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7698.144041001797,
      "totalMs": 17385.256081994157,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9992197751998901
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0005686170770786703
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00016254348156508058
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00004823402196052484
          },
          {
            "label": "The sound of a dog barking.",
            "score": 7.209198997770727e-7
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9992197751998901,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 9958.229334006319,
      "durationMs": 9962.55429199664,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.5020972938357143,
      "realTimeFactor": 1.9916458668012638,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06947697699069977,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7663.1660409912,
      "totalMs": 17625.72033298784,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9997052550315857
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00020508802845142782
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00004371672184788622
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00003144167203572579
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000014381625987880398
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9997052550315857,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 9616.597250002087,
      "durationMs": 9619.752332990174,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.5199344289893096,
      "realTimeFactor": 1.9233194500004174,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.16417637467384338,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7663.1660409912,
      "totalMs": 17282.918373981374,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9998252987861633
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00012907838390674442
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00004292851735954173
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000018302520174984238
          },
          {
            "label": "A person giving a speech.",
            "score": 9.836167009780183e-7
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9998252987861633,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 9583.06812500814,
      "durationMs": 9586.584874996333,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.521753569397249,
      "realTimeFactor": 1.916613625001628,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.11127196252346039,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7663.1660409912,
      "totalMs": 17249.750915987534,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9971395134925842,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9992963075637817
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00043248446309007704
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0001957510394277051
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00007384844502666965
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000016760664038883988
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9992963075637817
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 16368.054374994244,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "7aeb64c3f36edd59f472e11542309f21e6143beeae77aadbee34dec234d870b0",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "variation",
      "initNoiseLevel": 0.2,
      "peakNormalization": false,
      "inferenceMs": 2647.8418329934357,
      "durationMs": 2651.7725409939885,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 1.888330314030655,
      "realTimeFactor": 0.5295683665986871,
      "activationPlacement": {
        "audioEncoderOutput": "mps:0",
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.0705665796995163,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11500142592,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "totalMs": 19019.826915988233,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.996171772480011,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.999683141708374
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0001857801544247195
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0001037955516949296
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.000026187148250755854
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000010688427209970541
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.999683141708374
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 7731.958000003942,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "7aeb64c3f36edd59f472e11542309f21e6143beeae77aadbee34dec234d870b0",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "variation",
      "initNoiseLevel": 0.2,
      "peakNormalization": false,
      "inferenceMs": 15934.898666993831,
      "durationMs": 15938.061708991881,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.3137767051105613,
      "realTimeFactor": 3.186979733398766,
      "activationPlacement": {
        "audioEncoderOutput": "cpu",
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06243285909295082,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "totalMs": 23670.019708995824,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9998767971992493,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.99895179271698
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00045631214743480086
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0003860544820781797
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00020227623463142663
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000035849686810252024
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.99895179271698
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 16368.054374994244,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "7aeb64c3f36edd59f472e11542309f21e6143beeae77aadbee34dec234d870b0",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "inpaint",
      "editStartSeconds": 2,
      "editEndSeconds": 3,
      "peakNormalization": false,
      "inferenceMs": 2453.886707997299,
      "durationMs": 2457.4123329948634,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 2.0375838801786705,
      "realTimeFactor": 0.49077734159945974,
      "activationPlacement": {
        "audioEncoderOutput": "mps:0",
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.061997462064027786,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": 11500158976,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "totalMs": 18825.466707989108,
      "fallback": false,
      "grade": "fast"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9998508095741272,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9994373917579651
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0002785610849969089
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00018134269339498132
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.000101662844826933
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000001097448262044054
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9994373917579651
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 7731.958000003942,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "7aeb64c3f36edd59f472e11542309f21e6143beeae77aadbee34dec234d870b0",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "inpaint",
      "editStartSeconds": 2,
      "editEndSeconds": 3,
      "peakNormalization": false,
      "inferenceMs": 15349.102291002055,
      "durationMs": 15352.574207994621,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.32575194986687256,
      "realTimeFactor": 3.069820458200411,
      "activationPlacement": {
        "audioEncoderOutput": "cpu",
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06305339187383652,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 8,
      "cfgScale": 1,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "d4d3579395a8ac0df4ab13b5b1d7138ce7290f81",
      "totalMs": 23084.532207998564,
      "fallback": false,
      "grade": "slow"
    }
  }
]

Stable Audio 3.0 Medium Base

22 configurations tested
Passed22
Failed0
TestedNov 7, 2026
Runtimestable-audio-3 0.1.0+790793b

Quality criteria

No explicit quality gate recorded. Review per-case evidence; full certification is pending.

PrecisionDeviceResultPlatformInference / loadTask metric
FP32mpsAcceleratedPassedSlowmacos · aarch6416558 ms / 16005 ms3.312 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6416695 ms / 16005 ms3.339 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6416696 ms / 14475 ms3.339 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6416933 ms / 14475 ms3.387 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6417820 ms / 14475 ms3.564 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6418612 ms / 8505 ms3.722 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6418715 ms / 8505 ms3.743 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6418838 ms / 8357 ms3.768 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6418863 ms / 8505 ms3.773 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6419209 ms / 8357 ms3.842 RTF
FP32mpsAcceleratedPassedSlowmacos · aarch6420060 ms / 8357 ms4.012 RTF
FP32cpuPassedSlowmacos · aarch6445681 ms / 12915 ms9.136 RTF
FP32cpuPassedSlowmacos · aarch6445878 ms / 12915 ms9.176 RTF
FP32cpuPassedSlowmacos · aarch6446395 ms / 9559 ms9.279 RTF
FP32cpuPassedSlowmacos · aarch6446669 ms / 12915 ms9.334 RTF
FP32cpuPassedSlowmacos · aarch6447050 ms / 9559 ms9.410 RTF
FP32cpuPassedSlowmacos · aarch6447272 ms / 8112 ms9.454 RTF
FP32cpuPassedSlowmacos · aarch6447602 ms / 8112 ms9.520 RTF
FP32cpuPassedSlowmacos · aarch6447712 ms / 8112 ms9.542 RTF
FP32cpuPassedSlowmacos · aarch6448085 ms / 9559 ms9.617 RTF
FP32cpuPassedSlowmacos · aarch6452946 ms / 9096 ms10.589 RTF
FP32cpuPassedSlowmacos · aarch6454229 ms / 9096 ms10.846 RTF
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-07T14:56:07.531569+00:00
Environment
{ "os": "macos", "arch": "aarch64", "hardware": "Apple M3 Max", "osVersion": "27.2", "runtime": "Stable Audio 3 pinned 790793b / Torch 2.7.1 / Transformers 5.8.0", "interpreter": "App-owned CPython 3.13.18 private probe venv", "concurrentActivity": "Pinned candidate source downloads and tracked background archival service active. CLAP scoring runs between generation cases." }
Runtime
stable-audio-3 0.1.0+790793b
Artifact revision
c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c
Fixture
{
  "file": "model-samples/desktop/stable-audio3/medium-base/medium-base-piano.txt",
  "size": 83,
  "sha1": "bdf829476f6d8979181472b69c8ff1e40d822a37",
  "sha256": "598e2c9b870380f0e7632de52d63bcaf3ec5343badf49ba10b38b507d057d783",
  "url": "https://assets.localai.tools/samples/v1/files/598e2c9b870380f0e7632de52d63bcaf3ec5343badf49ba10b38b507d057d783/medium-base-piano.txt"
}
Rerun when
Checkpoint/runtime/adapter/config changes; Fixture or quality gate changes; Relevant device/backend behavior
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9993883371353149
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00028972470317967236
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00024293507158290595
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00007509159331675619
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000003922660198441008
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9993883371353149,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 17820.45579199621,
      "durationMs": 17825.910957995802,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.2805764374582201,
      "realTimeFactor": 3.564091158399242,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.13865958154201508,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11547131904,
      "loadMs": 14474.65750000265,
      "totalMs": 32300.56845799845,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9958729147911072
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0017352644354104996
          },
          {
            "label": "A slow piano melody.",
            "score": 0.0011222278699278831
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0010838531889021397
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00018579006427899003
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9958729147911072,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 16695.953500006,
      "durationMs": 16700.429791002534,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.29947376171107587,
      "realTimeFactor": 3.3391907000011996,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.3790886104106903,
      "clippedFraction": 0.006394557823129252,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11555520512,
      "loadMs": 14474.65750000265,
      "totalMs": 31175.087291005184,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9981200098991394
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0015578997554257512
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00029079720843583345
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000020730683900183067
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000010531459338380955
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9981200098991394,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 16932.913958997233,
      "durationMs": 16937.35687500157,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.2952829035869087,
      "realTimeFactor": 3.3865827917994467,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.15776193141937256,
      "clippedFraction": 0.00007482993197278911,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11555520512,
      "loadMs": 14474.65750000265,
      "totalMs": 31412.01437500422,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9993883371353149
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00028972470317967236
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00024293507158290595
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00007509159331675619
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000003922660198441008
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9993883371353149,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 18714.864334004233,
      "durationMs": 18721.44258399203,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.26716731207691313,
      "realTimeFactor": 3.7429728668008466,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.13865958154201508,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11547131904,
      "loadMs": 8504.585165996104,
      "totalMs": 27226.027749988134,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9958729147911072
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0017352644354104996
          },
          {
            "label": "A slow piano melody.",
            "score": 0.0011222278699278831
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0010838531889021397
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00018579006427899003
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9958729147911072,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 18611.632040992845,
      "durationMs": 18615.88737499551,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.2686491968564232,
      "realTimeFactor": 3.722326408198569,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.3790886104106903,
      "clippedFraction": 0.006394557823129252,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11555520512,
      "loadMs": 8504.585165996104,
      "totalMs": 27120.472540991614,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9981200098991394
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0015578997554257512
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00029079720843583345
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000020730683900183067
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000010531459338380955
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9981200098991394,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 18863.232582996716,
      "durationMs": 18867.0562079933,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.26506591476303965,
      "realTimeFactor": 3.7726465165993432,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.15776193141937256,
      "clippedFraction": 0.00007482993197278911,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11555520512,
      "loadMs": 8504.585165996104,
      "totalMs": 27371.641373989405,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9993883371353149
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00028972470317967236
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00024293507158290595
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00007509159331675619
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000003922660198441008
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9993883371353149,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 20060.47533400124,
      "durationMs": 20066.25799999165,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.24924633722538547,
      "realTimeFactor": 4.0120950668002475,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.13865958154201508,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11547131904,
      "loadMs": 8357.015707995743,
      "totalMs": 28423.27370798739,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9958729147911072
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0017352644354104996
          },
          {
            "label": "A slow piano melody.",
            "score": 0.0011222278699278831
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0010838531889021397
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00018579006427899003
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9958729147911072,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 18837.80270900752,
      "durationMs": 18841.491459010285,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.26542373743033154,
      "realTimeFactor": 3.7675605418015037,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.3790886104106903,
      "clippedFraction": 0.006394557823129252,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11555520512,
      "loadMs": 8357.015707995743,
      "totalMs": 27198.507167006028,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9981200098991394
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0015578997554257512
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00029079720843583345
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000020730683900183067
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000010531459338380955
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9981200098991394,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 19209.135957993567,
      "durationMs": 19212.78774998791,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.2602928112401293,
      "realTimeFactor": 3.841827191598713,
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.15776193141937256,
      "clippedFraction": 0.00007482993197278911,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11555520512,
      "loadMs": 8357.015707995743,
      "totalMs": 27569.803457983653,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9930558800697327
          },
          {
            "label": "A person giving a speech.",
            "score": 0.005351114086806774
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0011973599903285503
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0003485938941594213
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00004709307904704474
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9930558800697327,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 48084.93524999358,
      "durationMs": 48089.3437909981,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.10398267095515466,
      "realTimeFactor": 9.616987049998716,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.16359925270080566,
      "clippedFraction": 0.0008775510204081633,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 9558.986874995753,
      "totalMs": 57648.33066599385,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9982941746711731
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0007134299376048148
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0006548190140165389
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00022329027706291527
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00011419145448599011
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9982941746711731,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 46394.69662500778,
      "durationMs": 46398.16358400276,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.1077709385711316,
      "realTimeFactor": 9.278939325001556,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.41584402322769165,
      "clippedFraction": 0.009054421768707483,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 9558.986874995753,
      "totalMs": 55957.15045899851,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9906498789787292
          },
          {
            "label": "A slow piano melody.",
            "score": 0.008333015255630016
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0009985599899664521
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000094351707957685
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000009181627319776453
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9906498789787292,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 47050.40583301161,
      "durationMs": 47053.38770800154,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.10626900898040477,
      "realTimeFactor": 9.410081166602321,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.19958309829235077,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 9558.986874995753,
      "totalMs": 56612.37458299729,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9930558800697327
          },
          {
            "label": "A person giving a speech.",
            "score": 0.005351114086806774
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0011973599903285503
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0003485938941594213
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00004709307904704474
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9930558800697327,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 46669.388624999556,
      "durationMs": 46673.384666995844,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.1071366081132169,
      "realTimeFactor": 9.333877724999912,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.16359925270080566,
      "clippedFraction": 0.0008775510204081633,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 12915.188958009821,
      "totalMs": 59588.573625005665,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9982941746711731
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0007134299376048148
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0006548190140165389
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00022329027706291527
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00011419145448599011
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9982941746711731,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 45681.05916700733,
      "durationMs": 45684.27091700141,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.10945455493315702,
      "realTimeFactor": 9.136211833401466,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.41584402322769165,
      "clippedFraction": 0.009054421768707483,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 12915.188958009821,
      "totalMs": 58599.459875011235,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9906498789787292
          },
          {
            "label": "A slow piano melody.",
            "score": 0.008333015255630016
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0009985599899664521
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000094351707957685
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000009181627319776453
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9906498789787292,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 45878.079957998125,
      "durationMs": 45881.1080410087,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.1089845086057994,
      "realTimeFactor": 9.175615991599624,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.19958309829235077,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 12915.188958009821,
      "totalMs": 58796.29699901852,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9930558800697327
          },
          {
            "label": "A person giving a speech.",
            "score": 0.005351114086806774
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0011973599903285503
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0003485938941594213
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.00004709307904704474
          }
        ]
      },
      "expected": "A slow piano melody.",
      "score": 0.9930558800697327,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 47272.314417001326,
      "durationMs": 47276.47454199905,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.1057701545114484,
      "realTimeFactor": 9.454462883400264,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.16359925270080566,
      "clippedFraction": 0.0008775510204081633,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 8112.066292000236,
      "totalMs": 55388.54083399929,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.9982941746711731
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0007134299376048148
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0006548190140165389
          },
          {
            "label": "A slow piano melody.",
            "score": 0.00022329027706291527
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00011419145448599011
          }
        ]
      },
      "expected": "Electronic dance music with synthesizers and drums.",
      "score": 0.9982941746711731,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 47711.723915999755,
      "durationMs": 47714.94666599028,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.10479604570153225,
      "realTimeFactor": 9.542344783199951,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.41584402322769165,
      "clippedFraction": 0.009054421768707483,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 8112.066292000236,
      "totalMs": 55827.01295799052,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.9906498789787292
          },
          {
            "label": "A slow piano melody.",
            "score": 0.008333015255630016
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0009985599899664521
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000094351707957685
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000009181627319776453
          }
        ]
      },
      "expected": "A solo acoustic guitar playing a melody.",
      "score": 0.9906498789787292,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inferenceMs": 47601.85291701055,
      "durationMs": 47604.9049590074,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.1050379280133704,
      "realTimeFactor": 9.520370583402109,
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.19958309829235077,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 8112.066292000236,
      "totalMs": 55716.971251007635,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9998293519020081,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9986754059791565
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000882768421433866
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.0003410256467759609
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00008809144492261112
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000012580648217408452
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9986754059791565
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 16004.821875001653,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "e20fb135a1cd8df6f6ca3ae506504f70b979d8f3a596dad2b3e157fcbf9344a9",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "variation",
      "initNoiseLevel": 0.2,
      "peakNormalization": false,
      "inferenceMs": 16695.022542000515,
      "durationMs": 16699.4338749937,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.2994904611491986,
      "realTimeFactor": 3.339004508400103,
      "activationPlacement": {
        "audioEncoderOutput": "mps:0",
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.14078019559383392,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11555602432,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "totalMs": 32704.255749995355,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9994979500770569,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9962174296379089
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0032190585043281317
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00040540462941862643
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00013519938511308283
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000229192701226566
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9962174296379089
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 9096.144625000306,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "e20fb135a1cd8df6f6ca3ae506504f70b979d8f3a596dad2b3e157fcbf9344a9",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "variation",
      "initNoiseLevel": 0.2,
      "peakNormalization": false,
      "inferenceMs": 54229.44904099859,
      "durationMs": 54233.012999990024,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.09220082609026502,
      "realTimeFactor": 10.845889808199718,
      "activationPlacement": {
        "audioEncoderOutput": "cpu",
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.1392805129289627,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "totalMs": 63329.15762499033,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.999962568283081,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9986711740493774
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0009596028248779476
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.000270544522209093
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.0000888745635165833
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000009737564141687471
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9986711740493774
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 16004.821875001653,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 972,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "e20fb135a1cd8df6f6ca3ae506504f70b979d8f3a596dad2b3e157fcbf9344a9",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "inpaint",
      "editStartSeconds": 2,
      "editEndSeconds": 3,
      "peakNormalization": false,
      "inferenceMs": 16558.466917005717,
      "durationMs": 16563.011749996804,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.3019603218740588,
      "realTimeFactor": 3.3116933834011433,
      "activationPlacement": {
        "audioEncoderOutput": "mps:0",
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.14022332429885864,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 11555618816,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "totalMs": 32567.833624998457,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9999057650566101,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "A slow piano melody.",
            "score": 0.9987576007843018
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0008445489220321178
          },
          {
            "label": "A solo acoustic guitar playing a melody.",
            "score": 0.00031433458207175136
          },
          {
            "label": "Electronic dance music with synthesizers and drums.",
            "score": 0.00007671605271752924
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000068259778345236555
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9987576007843018
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 9096.144625000306,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 972,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "e20fb135a1cd8df6f6ca3ae506504f70b979d8f3a596dad2b3e157fcbf9344a9",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "inpaint",
      "editStartSeconds": 2,
      "editEndSeconds": 3,
      "peakNormalization": false,
      "inferenceMs": 52946.061332986574,
      "durationMs": 52949.16624999314,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.09443573089514949,
      "realTimeFactor": 10.589212266597315,
      "activationPlacement": {
        "audioEncoderOutput": "cpu",
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.13964521884918213,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "c397d9a6912dc8b1f71a80e7b9673f4d307e7f8c",
      "totalMs": 62045.31087499345,
      "fallback": false,
      "grade": "slow"
    }
  }
]

Stable Audio 3.0 Small SFX Base

22 configurations tested
Passed22
Failed0
TestedNov 7, 2026
Runtimestable-audio-3 0.1.0+790793b

Quality criteria

No explicit quality gate recorded. Review per-case evidence; full certification is pending.

PrecisionDeviceResultPlatformInference / loadTask metric
FP32mpsAcceleratedPassedFastmacos · aarch646018 ms / 3745 ms1.204 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646094 ms / 3527 ms1.219 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646129 ms / 4729 ms1.226 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646130 ms / 3745 ms1.226 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646158 ms / 4729 ms1.232 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646171 ms / 3527 ms1.234 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646203 ms / 3595 ms1.241 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646215 ms / 3595 ms1.243 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646627 ms / 3745 ms1.325 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646806 ms / 3595 ms1.361 RTF
FP32mpsAcceleratedPassedFastmacos · aarch646960 ms / 4729 ms1.392 RTF
FP32cpuPassedSlowmacos · aarch6415449 ms / 7275 ms3.090 RTF
FP32cpuPassedSlowmacos · aarch6415494 ms / 3058 ms3.099 RTF
FP32cpuPassedSlowmacos · aarch6415543 ms / 3058 ms3.109 RTF
FP32cpuPassedSlowmacos · aarch6415657 ms / 3166 ms3.131 RTF
FP32cpuPassedSlowmacos · aarch6415665 ms / 7275 ms3.133 RTF
FP32cpuPassedSlowmacos · aarch6415665 ms / 7275 ms3.133 RTF
FP32cpuPassedSlowmacos · aarch6415802 ms / 3954 ms3.160 RTF
FP32cpuPassedSlowmacos · aarch6415859 ms / 3954 ms3.172 RTF
FP32cpuPassedSlowmacos · aarch6416119 ms / 3954 ms3.224 RTF
FP32cpuPassedSlowmacos · aarch6416268 ms / 3166 ms3.254 RTF
FP32cpuPassedSlowmacos · aarch6416379 ms / 3166 ms3.276 RTF
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-07T14:48:50.948184+00:00
Environment
{ "os": "macos", "arch": "aarch64", "hardware": "Apple M3 Max", "osVersion": "27.2", "runtime": "Stable Audio 3 pinned 790793b / Torch 2.7.1 / Transformers 5.8.0", "interpreter": "App-owned CPython 3.13.18 private probe venv", "concurrentActivity": "Pinned candidate source downloads and tracked background archival service active. CLAP scoring runs between generation cases." }
Runtime
stable-audio-3 0.1.0+790793b
Artifact revision
b332479a012f83c24c4381f331ce594284ec8a81
Fixture
{
  "file": "model-samples/desktop/stable-audio3/small-sfx-base/small-sfx-base-dog.txt",
  "size": 45,
  "sha1": "c281230c4a32a4bd334fdffb323aa72a51bd19bb",
  "sha256": "9fe69144aba90458a13e93766c71c045a568e44bbf973df0dc9dc96af91cb29b",
  "url": "https://assets.localai.tools/samples/v1/files/9fe69144aba90458a13e93766c71c045a568e44bbf973df0dc9dc96af91cb29b/small-sfx-base-dog.txt"
}
Rerun when
Checkpoint/runtime/adapter/config changes; Fixture or quality gate changes; Relevant device/backend behavior
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9999579191207886
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.000030061137294978835
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000006071945790608879
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000036917147099302383
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.0000022056515263102483
          }
        ]
      },
      "expected": "The sound of a dog barking.",
      "score": 0.9999579191207886,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.8059592247009277,
      "normalizationDivisor": 2.8632235527038574,
      "inferenceMs": 6959.96408299834,
      "durationMs": 6964.285791007569,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.7183945118644935,
      "realTimeFactor": 1.391992816599668,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.07176990061998367,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 4728.543541990803,
      "totalMs": 11692.829332998372,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.867167055606842
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.11554811149835587
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.006187716498970985
          },
          {
            "label": "A person giving a speech.",
            "score": 0.005959852132946253
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.005137272644788027
          }
        ]
      },
      "expected": "The sound of rain falling and water splashing.",
      "score": 0.867167055606842,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 0.5267046689987183,
      "normalizationDivisor": 1,
      "inferenceMs": 6129.183250013739,
      "durationMs": 6132.635459012818,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.8157693767744327,
      "realTimeFactor": 1.2258366500027478,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.04173696041107178,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 4728.543541990803,
      "totalMs": 10861.17900100362,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.9999692440032959
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.000016338593923137523
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000009873332601273432
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.000002867002422135556
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000001662073145780596
          }
        ]
      },
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "score": 0.9999692440032959,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.1252286434173584,
      "normalizationDivisor": 2.168600559234619,
      "inferenceMs": 6158.491582988063,
      "durationMs": 6162.008749990491,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.8118871208352014,
      "realTimeFactor": 1.2316983165976125,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.022972701117396355,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 4728.543541990803,
      "totalMs": 10890.552291981294,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9999579191207886
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.000030061137294978835
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000006071945790608879
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000036917147099302383
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.0000022056515263102483
          }
        ]
      },
      "expected": "The sound of a dog barking.",
      "score": 0.9999579191207886,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.8059592247009277,
      "normalizationDivisor": 2.8632235527038574,
      "inferenceMs": 6626.949917001184,
      "durationMs": 6631.587833006051,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.7544949128365515,
      "realTimeFactor": 1.3253899834002367,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.07176990061998367,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 3745.4033750109375,
      "totalMs": 10376.991208016989,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.867167055606842
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.11554811149835587
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.006187716498970985
          },
          {
            "label": "A person giving a speech.",
            "score": 0.005959852132946253
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.005137272644788027
          }
        ]
      },
      "expected": "The sound of rain falling and water splashing.",
      "score": 0.867167055606842,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 0.5267046689987183,
      "normalizationDivisor": 1,
      "inferenceMs": 6129.9000840081135,
      "durationMs": 6133.32012500905,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.8156739802405859,
      "realTimeFactor": 1.2259800168016226,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.04173696041107178,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 3745.4033750109375,
      "totalMs": 9878.723500019987,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.9999692440032959
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.000016338593923137523
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000009873332601273432
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.000002867002422135556
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000001662073145780596
          }
        ]
      },
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "score": 0.9999692440032959,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.1252286434173584,
      "normalizationDivisor": 2.168600559234619,
      "inferenceMs": 6018.242624995764,
      "durationMs": 6021.843333001016,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.8308073156162459,
      "realTimeFactor": 1.2036485249991529,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.022972701117396355,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 3745.4033750109375,
      "totalMs": 9767.246708011953,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9999579191207886
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.000030061137294978835
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000006071945790608879
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000036917147099302383
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.0000022056515263102483
          }
        ]
      },
      "expected": "The sound of a dog barking.",
      "score": 0.9999579191207886,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.8059592247009277,
      "normalizationDivisor": 2.8632235527038574,
      "inferenceMs": 6805.893416996696,
      "durationMs": 6811.000707995845,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.7346574055234617,
      "realTimeFactor": 1.3611786833993391,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.07176990061998367,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 3594.905291000032,
      "totalMs": 10405.905998995877,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.867167055606842
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.11554811149835587
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.006187716498970985
          },
          {
            "label": "A person giving a speech.",
            "score": 0.005959852132946253
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.005137272644788027
          }
        ]
      },
      "expected": "The sound of rain falling and water splashing.",
      "score": 0.867167055606842,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 0.5267046689987183,
      "normalizationDivisor": 1,
      "inferenceMs": 6214.838999992935,
      "durationMs": 6219.12408299977,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.8045260705877794,
      "realTimeFactor": 1.242967799998587,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.04173696041107178,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 3594.905291000032,
      "totalMs": 9814.029373999801,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.9999692440032959
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.000016338593923137523
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000009873332601273432
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.000002867002422135556
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.000001662073145780596
          }
        ]
      },
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "score": 0.9999692440032959,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.1252286434173584,
      "normalizationDivisor": 2.168600559234619,
      "inferenceMs": 6203.088291993481,
      "durationMs": 6206.5794589871075,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.8060501099837085,
      "realTimeFactor": 1.2406176583986963,
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.022972701117396355,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405381632,
      "loadMs": 3594.905291000032,
      "totalMs": 9801.48474998714,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "mps",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9998403787612915
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.00010655705409590155
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000038982580008450896
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.000011213009202037938
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000028017796012136387
          }
        ]
      },
      "expected": "The sound of a dog barking.",
      "score": 0.9998403787612915,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.9340882301330566,
      "normalizationDivisor": 2.9939675331115723,
      "inferenceMs": 15657.022249986767,
      "durationMs": 15661.137874994893,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.3193455256157809,
      "realTimeFactor": 3.1314044499973535,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06738532334566116,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 3166.2454170000274,
      "totalMs": 18827.38329199492,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.881836473941803
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.10592535138130188
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0053984010592103004
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.004171841312199831
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.002667957218363881
          }
        ]
      },
      "expected": "The sound of rain falling and water splashing.",
      "score": 0.881836473941803,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 0.5314595699310303,
      "normalizationDivisor": 1,
      "inferenceMs": 16379.003082998679,
      "durationMs": 16382.867332999012,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.30526888447746703,
      "realTimeFactor": 3.275800616599736,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.03846383094787598,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 3166.2454170000274,
      "totalMs": 19549.11274999904,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.9997418522834778
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.0001880517665995285
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000560919834242668
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000012176455129520036
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000017438064787711482
          }
        ]
      },
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "score": 0.9997418522834778,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 1.409977674484253,
      "normalizationDivisor": 1.4387526512145996,
      "inferenceMs": 16267.59299999685,
      "durationMs": 16270.660624999437,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.30735954606197535,
      "realTimeFactor": 3.25351859999937,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06546288728713989,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 3166.2454170000274,
      "totalMs": 19436.906041999464,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9998403787612915
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.00010655705409590155
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000038982580008450896
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.000011213009202037938
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000028017796012136387
          }
        ]
      },
      "expected": "The sound of a dog barking.",
      "score": 0.9998403787612915,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.9340882301330566,
      "normalizationDivisor": 2.9939675331115723,
      "inferenceMs": 15858.79129100067,
      "durationMs": 15863.135457999306,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.31528254002796113,
      "realTimeFactor": 3.1717582582001342,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06738532334566116,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 3954.268667002907,
      "totalMs": 19817.404125002213,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.881836473941803
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.10592535138130188
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0053984010592103004
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.004171841312199831
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.002667957218363881
          }
        ]
      },
      "expected": "The sound of rain falling and water splashing.",
      "score": 0.881836473941803,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 0.5314595699310303,
      "normalizationDivisor": 1,
      "inferenceMs": 15801.822333000018,
      "durationMs": 15805.834208003944,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.3164192011929004,
      "realTimeFactor": 3.1603644666000035,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.03846383094787598,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 3954.268667002907,
      "totalMs": 19760.10287500685,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.9997418522834778
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.0001880517665995285
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000560919834242668
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000012176455129520036
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000017438064787711482
          }
        ]
      },
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "score": 0.9997418522834778,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 1.409977674484253,
      "normalizationDivisor": 1.4387526512145996,
      "inferenceMs": 16118.524249992333,
      "durationMs": 16121.638707991224,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.3102020955797103,
      "realTimeFactor": 3.223704849998467,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06546288728713989,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 3954.268667002907,
      "totalMs": 20075.90737499413,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9998403787612915
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.00010655705409590155
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000038982580008450896
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.000011213009202037938
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000028017796012136387
          }
        ]
      },
      "expected": "The sound of a dog barking.",
      "score": 0.9998403787612915,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 2.9340882301330566,
      "normalizationDivisor": 2.9939675331115723,
      "inferenceMs": 15664.661583999987,
      "durationMs": 15668.866459003766,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.319189787355958,
      "realTimeFactor": 3.132932316799997,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06738532334566116,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7274.686541000847,
      "totalMs": 22943.553000004613,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.881836473941803
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.10592535138130188
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0053984010592103004
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.004171841312199831
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.002667957218363881
          }
        ]
      },
      "expected": "The sound of rain falling and water splashing.",
      "score": 0.881836473941803,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 0.5314595699310303,
      "normalizationDivisor": 1,
      "inferenceMs": 15665.423917002045,
      "durationMs": 15668.854000003194,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.3191742544913442,
      "realTimeFactor": 3.1330847834004087,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.03846383094787598,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7274.686541000847,
      "totalMs": 22943.54054100404,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.9997418522834778
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.0001880517665995285
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000560919834242668
          },
          {
            "label": "A person giving a speech.",
            "score": 0.000012176455129520036
          },
          {
            "label": "The sound of a dog barking.",
            "score": 0.0000017438064787711482
          }
        ]
      },
      "expected": "The sound of a wooden door creaking and slamming shut.",
      "score": 0.9997418522834778,
      "metric": "CLAP top-1 intended class among five frozen labels plus signal validity"
    },
    "runtime": {
      "inputMode": "text",
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 1.409977674484253,
      "normalizationDivisor": 1.4387526512145996,
      "inferenceMs": 15448.636083005113,
      "durationMs": 15451.702125006705,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.32365316738222927,
      "realTimeFactor": 3.0897272166010223,
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "activationPlacement": {
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06546288728713989,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "loadMs": 7274.686541000847,
      "totalMs": 22726.38866600755,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "precision": "fp32",
      "device": "cpu",
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "cpuThreads": 4,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9978880286216736,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9995604157447815
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.00034680141834542155
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00006633648445131257
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.000015136376532609574
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.000011375642316124868
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9995604157447815
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3526.515499994275,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "68c4a088673ab70321d273c901c2064868accade24fdf0d2f294ecc914f96b33",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "variation",
      "initNoiseLevel": 0.2,
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 0.9857620596885681,
      "normalizationDivisor": 1.0058796405792236,
      "inferenceMs": 6094.372209001449,
      "durationMs": 6097.775875008665,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.820429049708344,
      "realTimeFactor": 1.2188744418002897,
      "activationPlacement": {
        "audioEncoderOutput": "mps:0",
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.07372812926769257,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405447168,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "totalMs": 9624.29137500294,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9974175095558167,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9999517202377319
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.000042921867134282365
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0000032140358143806225
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.000001240123651768954
          },
          {
            "label": "Music with instruments and singing.",
            "score": 9.167787311525899e-7
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9999517202377319
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3058.2683329994325,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "68c4a088673ab70321d273c901c2064868accade24fdf0d2f294ecc914f96b33",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "variation",
      "initNoiseLevel": 0.2,
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 0.9713859558105469,
      "normalizationDivisor": 1,
      "inferenceMs": 15493.719666003017,
      "durationMs": 15496.928125008708,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.32271140228328865,
      "realTimeFactor": 3.0987439332006033,
      "activationPlacement": {
        "audioEncoderOutput": "cpu",
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.07010193169116974,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "totalMs": 18555.19645800814,
      "fallback": false,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9990456700325012,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9999387264251709
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.00004759976582136005
          },
          {
            "label": "A person giving a speech.",
            "score": 0.00000546487353858538
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.00000428022576670628
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.00000394359130950761
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9999387264251709
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3526.515499994275,
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "mps:0/torch.float32": 672,
        "mps:0/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "68c4a088673ab70321d273c901c2064868accade24fdf0d2f294ecc914f96b33",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "inpaint",
      "editStartSeconds": 2,
      "editEndSeconds": 3,
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 1.5664774179458618,
      "normalizationDivisor": 1.5984463691711426,
      "inferenceMs": 6171.133125011693,
      "durationMs": 6174.510083001223,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.8102239732497306,
      "realTimeFactor": 1.2342266250023386,
      "activationPlacement": {
        "audioEncoderOutput": "mps:0",
        "diffusionOutput": "mps:0"
      },
      "waveformDevice": "mps:0",
      "rms": 0.05912255495786667,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": 3405463552,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "totalMs": 9701.025582995499,
      "fallback": false,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "signal": true,
      "semantic": true,
      "changed": true,
      "outsideEnvelopeCosine": 0.9990807771682739,
      "preserved": true,
      "clap": {
        "predictions": [
          {
            "label": "The sound of a dog barking.",
            "score": 0.9999561309814453
          },
          {
            "label": "The sound of a wooden door creaking and slamming shut.",
            "score": 0.00003752804332179949
          },
          {
            "label": "A person giving a speech.",
            "score": 0.0000026943407647195272
          },
          {
            "label": "Music with instruments and singing.",
            "score": 0.0000018522435993872932
          },
          {
            "label": "The sound of rain falling and water splashing.",
            "score": 0.0000016697679257049458
          }
        ]
      },
      "metric": "CLAP intended source class, valid signal, actual reference encoder execution and source dependence; editing also checks outside-interval amplitude envelope",
      "score": 0.9999561309814453
    },
    "runtime": {
      "engine": "stable-audio-3",
      "version": "0.1.0+790793b",
      "torchVersion": "2.7.1",
      "loadMs": 3058.2683329994325,
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "cpu/torch.float32": 672,
        "cpu/torch.bfloat16": 134
      },
      "accelerationDetails": "Neural text encoder, diffusion and audio decoder on selected device; tokenization and WAV encoding on CPU",
      "inputAudioSha256": "68c4a088673ab70321d273c901c2064868accade24fdf0d2f294ecc914f96b33",
      "inputAudioSeconds": 5,
      "inputSampleRate": 44100,
      "inputMode": "inpaint",
      "editStartSeconds": 2,
      "editEndSeconds": 3,
      "peakNormalization": true,
      "decoderPeakBeforeNormalization": 1.3293513059616089,
      "normalizationDivisor": 1.3564809560775757,
      "inferenceMs": 15542.807208010345,
      "durationMs": 15545.916125003714,
      "audioSeconds": 5,
      "audioSecondsPerSecond": 0.3216922099775602,
      "realTimeFactor": 3.108561441602069,
      "activationPlacement": {
        "audioEncoderOutput": "cpu",
        "diffusionOutput": "cpu"
      },
      "waveformDevice": "cpu",
      "rms": 0.06496044993400574,
      "clippedFraction": 0,
      "seed": 42,
      "steps": 50,
      "cfgScale": 7,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "artifactRevision": "b332479a012f83c24c4381f331ce594284ec8a81",
      "totalMs": 18604.184458003147,
      "fallback": false,
      "grade": "slow"
    }
  }
]

Something not quite right?

Tell us what happened with Stable Audio 3.0. Loading, results, or the page itself—we’re here for all of it.

Attachments Optional

Up to 3 files · 50 MB each

Browser, OS, CPU/GPU, memory, screen, storage, and network details are included automatically. Model inputs, results, and browsing history are not collected.