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.

Wav2Vec2 Base 960h — Word Timestamps

Transcribe English speech locally and return word-level timestamps using actual CTC character emission frames. Uses the 94M-parameter Wav2Vec2 acoustic model; no speaker labeling or translation.

Browser inferencePowered by WebAI · Private · Unlimited runs
About this model

Transcribe English speech locally and return word-level timestamps using actual CTC character emission frames. Uses the 94M-parameter Wav2Vec2 acoustic model; no speaker labeling or translation.

Created by
Meta
License
Apache-2.0
Model released
Unknown
Source repository created
Mar 2, 2022
Model weights · FP32
377.9 MB

Best for

  • English caption/search prototypes with estimated word timings
  • Local comparison of CTC ASR and word-timing behavior

Keep in mind

  • English only; trained on LibriSpeech read speech. Accents, conversational audio, noise, music and overlapping speakers can cause recognition and timing errors.
  • Word times are acoustic emission envelopes at20ms frame stride, not human-annotated boundaries or forced alignment. CTC character spikes can be shorter than spoken phonemes.
  • No language model, beam search, punctuation restoration, diarization or translation; output normally uses uppercase English.
  • Mono16kHz Float32Array PCM required, minimum400samples (25ms receptive field). Exact digital silence returns an empty result; quiet/non-speech audio may still hallucinate.
  • Whole-input inference; no arbitrary duration cap, but attention and decoding memory grow rapidly with duration. Longer recordings can exhaust browser/GPU memory.
  • Timestamp tests cover finite/bounded/monotonic word records, reconstruction, known leading-silence shift and utterance boundaries; they do not establish universal word-boundary accuracy.
  • Per-word confidence is a mean of CTC character softmax scores; it is not calibrated confidence that the word or time is correct.

Processed on your deviceInference runs in this browser. Input/output sharing requires consent for each run. Basic run logs include device information and approximate connection location. Model and runtime files download from external hosts. Saved downloads may be cleared by your browser.

Sources & technical details
Original model
facebook/wav2vec2-base-960h
Release verification
Unknown
Release evidence
Exact original checkpoint public release date has not been verified. Repository creation, training, paper, artifact conversion and update dates are not release evidence.
Source updated
Nov 14, 2022
Browser model files
Xenova/wav2vec2-base-960h
Files updated
Aug 18, 2025
Artifact revision
a19f851b3d42865797e410752b4c570c871e4825

Input

AUDIO

Sample result

Wav2Vec2 Base 960h — Word Timestamps
Record a clip, then run inference. This does not send live audio to the model.

Output

Your transcript or answer will appear here

Choose the input task, then add a WAV clip or text, or try a named sample.

Benchmark evidence

Saved worker evidence is separate from benchmarks run on your device. It does not guarantee speed or quality on every browser.

Wav2Vec2 Base 960h — Word Timestamps

19 configurations tested
Passed16
Failed3
TestedNov 2, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

{ "maximumWordErrorRate": 0.35, "timestamps": "Complete word/chunk reconstruction, finite/bounded/ordered word times, exact silence negative;0.5s prefix median word-start shift error<=0.15s and >=70% same-index matched words", "scope": "JFK and shared conversational fixtures; English-only consistency, not human-aligned timing certification" }

PrecisionDeviceResult
FP16GPU · WebGPUAcceleratedPassedVery fast
FP32GPU · WebGPUAcceleratedPassedVery fast
Q4GPU · WebGPUAcceleratedPassedVery fast
FP16GPU · WebGPUAcceleratedPassedVery fast
FP32GPU · WebGPUAcceleratedPassedVery fast
Q4GPU · WebGPUAcceleratedPassedVery fast
FP32GPU · WebGPUAcceleratedPassedVery fast
FP32GPU · WebGPUAcceleratedPassedVery fast
Q4GPU · WebGPUAcceleratedPassedVery fast
FP16GPU · WebGPUAcceleratedPassedVery fast
Q4CPU · WASMPassedFast
FP32CPU · WASMPassedFast
Q4CPU · WASMPassedFast
FP32CPU · WASMPassedFast
Q4CPU · WASMPassedFast
FP32CPU · WASMPassedFast
FP16CPU · WASMFailed—
Q8CPU · WASMFailed—
Q8GPU · WebGPUFailed—
Test setup & sample details
Quality gate
{ "maximumWordErrorRate": 0.35, "timestamps": "Complete word/chunk reconstruction, finite/bounded/ordered word times, exact silence negative;0.5s prefix median word-start shift error<=0.15s and >=70% same-index matched words", "scope": "JFK and shared conversational fixtures; English-only consistency, not human-aligned timing certification" }
Tested
2026-11-02T13:38:44.530Z
Environment
Chrome156.0.8078.79; Apple M3 Max; Darwin27.2.0 arm64; separate disposable persistent profile per precision/device, module workers, pinned remote HF artifacts
Runtime
Transformers.js 4.2.0
Artifact revision
a19f851b3d42865797e410752b4c570c871e4825
Fixture
{
  "file": "model-samples/vosk-jfk.flac",
  "url": "https://assets.localai.tools/assets/benchmarks/vosk/63a4b1e4c1dc655ac70961ffbf518acd249df237e5a0152faae9a4a836949715/vosk-jfk.flac",
  "size": 1152693,
  "sha1": "537ddf75dff344f8d31c1cd6ebee23ee8362d921",
  "sha256": "63a4b1e4c1dc655ac70961ffbf518acd249df237e5a0152faae9a4a836949715"
}
Rerun when
Artifact/runtime/CTC preprocessing-decoding changes; Fixture/config/quality/timestamp gates change; Relevant browser/backend behavior changes
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22
    },
    "runtime": {
      "loadMs": 44015.69999998808,
      "inferenceMs": 3738.699999988079,
      "totalMs": 47754.39999997616,
      "realTimeFactor": 0.3398818181807345,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22,
      "leadingSilenceShift": {
        "seconds": 0.5,
        "alignedWords": 22,
        "medianAbsoluteShiftErrorSeconds": 0,
        "maximumMedianErrorSeconds": 0.15
      }
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 3808.0999999940395,
      "totalMs": 3808.0999999940395,
      "realTimeFactor": 0.3311391304342643,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.30303030303030304,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 66
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 10888.09999999404,
      "totalMs": 10888.09999999404,
      "realTimeFactor": 0.36918512490549343,
      "grade": "good"
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22
    },
    "runtime": {
      "loadMs": 44770.40000000596,
      "inferenceMs": 174.69999998807907,
      "totalMs": 44945.09999999404,
      "realTimeFactor": 0.01588181818073446,
      "grade": "excellent"
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22,
      "leadingSilenceShift": {
        "seconds": 0.5,
        "alignedWords": 22,
        "medianAbsoluteShiftErrorSeconds": 0,
        "maximumMedianErrorSeconds": 0.15
      }
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 114.59999999403954,
      "totalMs": 114.59999999403954,
      "realTimeFactor": 0.009965217390786046,
      "grade": "excellent"
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.30303030303030304,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 66
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 373,
      "totalMs": 373,
      "realTimeFactor": 0.01264739041612627,
      "grade": "excellent"
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 25379.09999999404,
      "grade": "failed"
    },
    "failure": "Error: [WorkerError:runtime_error] init: Can't create a session. ERROR_CODE: 1, ERROR_MESSAGE: /mnt/vss/_work/1/s/onnxruntime/core/graph/graph_utils.cc:30 int onnxruntime::graph_utils::GetIndexFromName(const Node &, const std::string &, bool) itr != node_args.end() was false. Attempting to get index by a name which does not exist:InsertedPrecisionFreeCast_/wav2vec2/encoder/layers.11/layer_norm/Constant_output_0for node: /wav2vec2/feature_projection/layer_norm/Mul/SimplifiedLayerNormFusion/\n"
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22
    },
    "runtime": {
      "loadMs": 25708.70000000298,
      "inferenceMs": 829,
      "totalMs": 26537.70000000298,
      "realTimeFactor": 0.07536363636363637,
      "grade": "excellent"
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22,
      "leadingSilenceShift": {
        "seconds": 0.5,
        "alignedWords": 22,
        "medianAbsoluteShiftErrorSeconds": 0,
        "maximumMedianErrorSeconds": 0.15
      }
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 95.1000000089407,
      "totalMs": 95.1000000089407,
      "realTimeFactor": 0.008269565218168757,
      "grade": "excellent"
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.25757575757575757,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 70
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 307,
      "totalMs": 307,
      "realTimeFactor": 0.01040951436394307,
      "grade": "excellent"
    }
  },
  {
    "precision": "q8",
    "device": "wasm",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": null,
      "grade": "failed"
    },
    "failure": "Failed transcription/timestamp quality gate"
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": null,
      "grade": "failed"
    },
    "failure": "Failed transcription/timestamp quality gate"
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.09090909090909091,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22
    },
    "runtime": {
      "loadMs": 14773,
      "inferenceMs": 3684.7999999970198,
      "totalMs": 18457.79999999702,
      "realTimeFactor": 0.33498181818154726,
      "grade": "good"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.09090909090909091,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22,
      "leadingSilenceShift": {
        "seconds": 0.5,
        "alignedWords": 22,
        "medianAbsoluteShiftErrorSeconds": 0,
        "maximumMedianErrorSeconds": 0.15
      }
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 3750.2000000029802,
      "totalMs": 3750.2000000029802,
      "realTimeFactor": 0.3261043478263461,
      "grade": "good"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.25757575757575757,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 67
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 10734.09999999404,
      "totalMs": 10734.09999999404,
      "realTimeFactor": 0.363963414117066,
      "grade": "good"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.09090909090909091,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22
    },
    "runtime": {
      "loadMs": 16914.5,
      "inferenceMs": 235.5,
      "totalMs": 17150,
      "realTimeFactor": 0.02140909090909091,
      "grade": "excellent"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.09090909090909091,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22,
      "leadingSilenceShift": {
        "seconds": 0.5,
        "alignedWords": 22,
        "medianAbsoluteShiftErrorSeconds": 0,
        "maximumMedianErrorSeconds": 0.15
      }
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 115.29999999701977,
      "totalMs": 115.29999999701977,
      "realTimeFactor": 0.010026086956262588,
      "grade": "excellent"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0.25757575757575757,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 67
    },
    "runtime": {
      "loadMs": 0,
      "inferenceMs": 377.20000000298023,
      "totalMs": 377.20000000298023,
      "realTimeFactor": 0.012789800710457162,
      "grade": "excellent"
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "wordErrorRate": 0,
      "maximumWordErrorRate": 0.35,
      "validWordTimes": true,
      "completeReconstruction": true,
      "wordCount": 22
    },
    "runtime": {
      "loadMs": 45783.19999998808,
      "inferenceMs": 171.70000000298023,
      "totalMs": 45954.89999999106,
      "realTimeFactor": 0.01560909090936184,
      "grade": "excellent"
    }
  }
]

Something not quite right?

Tell us what happened with Wav2Vec2 ASR. 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.