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.

Hunyuan3D-2

Generate an untextured 3D mesh from an isolated object image using Hunyuan3D-2 Mini Turbo. Returns a GLB file for 3D tools.

Desktop inferencePrivate · Runs on your computer
About this model

Generate an untextured 3D mesh from an isolated object image using Hunyuan3D-2 Mini Turbo. Returns a GLB file for 3D tools.

Created by
Tencent Hunyuan
Model released
Mar 19, 2025
Source repository created
Jan 21, 2025
Model weights · FP32
3.8 GB

Best for

  • Local prototype meshes from authorized isolated-object images
  • Coarse shape exploration and GLB export

Keep in mind

  • Shape only; texture, materials, animation and multiview generation are not implemented.
  • Single-view geometry is an estimate; unseen surfaces, thin details, holes and topology can be incorrect. Review meshes before production, fabrication or simulation.
  • Use a clear isolated object, preferably transparent PNG. No automatic background removal or content safety classifier is included.
  • Quality evidence covers authored sphere/cube fixtures and a coarse 64-cell extraction grid, not arbitrary-object fidelity or manufacturing suitability.
  • Mini Turbo has a 0.6B shape DiT plus a larger DINOv2 encoder and VAE; the complete download and memory use exceed 0.6B alone.
  • Four steps and resolution 64 are tested. Higher resolutions require substantially more time and memory; they do not imply validated quality.
  • Only the recorded Apple Silicon Mac CPU/Metal configuration is verified; Windows, Intel Mac and other Mac versions/chips remain unverified.
  • GLB is complete; optional SVG is an orthographic preview with at most 12000 sampled triangles, not a substitute for the native mesh.
  • Hunyuan3D-2.1 is a newer separate family; this adapter intentionally implements the requested Hunyuan3D-2 Mini Turbo checkpoint.

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://github.com/Tencent-Hunyuan/Hunyuan3D-2
Release verification
Verified
Release evidence
Official model zoo explicitly dates Mini Turbo March 19, 2025; standard 2.0 and Mini have distinct earlier release dates.
Source updated
Oct 28, 2025
Native model files
https://huggingface.co/tencent/Hunyuan3D-2mini
Files updated
Oct 17, 2025
Artifact revision
f90a0f7df7d5e6f71109cf333f6a95a0ae3194a6

Desktop app

Sample result

Hunyuan3D-2 Mini Turbo — Image to 3D

Run Hunyuan3D-2 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 / FP32Tested32.0 GiBRAM estimate
macOS Apple SiliconMPS / FP32Tested56.0 GiBRAM estimate
Windows Intel / AMDCPU / FP32Unverified32.0 GiBRAM estimate
macOS Intel / AMDCPU / FP32Unverified32.0 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.

Hunyuan3D-2 Mini Turbo — Image to 3D

6 configurations tested
Passed6
Failed0
TestedNov 3, 2026
Runtimehy3dgen f8db63096c8282cb27354314d896feba5ba6ff8a

Quality criteria

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

PrecisionDeviceResultPlatformInference / load
FP32mpsAcceleratedPassedModeratemacos · aarch6435433 ms / 0 ms
FP32mpsAcceleratedPassedModeratemacos · aarch6436272 ms / 13270 ms
FP32mpsAcceleratedPassedModeratemacos · aarch6436888 ms / 0 ms
FP32cpuPassedModeratemacos · aarch64103621 ms / 0 ms
FP32cpuPassedModeratemacos · aarch64103869 ms / 12703 ms
FP32cpuPassedModeratemacos · aarch64104009 ms / 0 ms
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-03T19:42:55.402568+00:00
Environment
{ "hardware": "Apple M3 Max", "os": "macOS 27.2", "python": "App-owned CPython 3.13.18" }
Runtime
hy3dgen f8db63096c8282cb27354314d896feba5ba6ff8a
Artifact revision
f90a0f7df7d5e6f71109cf333f6a95a0ae3194a6
Fixture
{
  "file": "model-samples/desktop/hunyuan3d/sphere-input.png",
  "url": "https://assets.localai.tools/samples/v1/files/88d5f154e1d161e447749e5c0486d208d812a7234169e5447edd011a68236155/sphere-input.png",
  "size": 56359,
  "sha1": "832fc2842dab93d3d05eb6bfcccbb3fa91a2c76b",
  "sha256": "88d5f154e1d161e447749e5c0486d208d812a7234169e5447edd011a68236155"
}
Rerun when
Checkpoint revision/files/checksums; Reviewed upstream implementation/runtime versions; Fixtures or quality criteria; Device/backend/precision or driver behavior; Preprocessing, extraction resolution, steps or visualization logic
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "Analytic primitive surface error, finite topology, closed mesh and nondegenerate bounds",
      "score": 0.99262621481631,
      "threshold": 0.85,
      "meanSurfaceError": 0.0073737851836900535,
      "watertight": true,
      "scope": "Sphere/cube procedural pilot; not arbitrary-object or texture certification"
    },
    "runtime": {
      "engine": "hy3dgen",
      "version": "f8db63096c8282cb27354314d896feba5ba6ff8a",
      "torchVersion": "2.6.0",
      "transformersVersion": "4.51.3",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "model": [
          "mps:0/torch.float32"
        ],
        "vae": [
          "mps:0/torch.float32"
        ],
        "conditioner": [
          "mps:0/torch.float32"
        ]
      },
      "loadMs": 13270.460375002585,
      "accelerationDetails": "Shape DiT, VAE and DINOv2 on selected device; preprocessing, marching cubes and GLB encoding on CPU. No texture generation.",
      "inferenceMs": 36272.35441701487,
      "durationMs": 36312.1534170059,
      "meshesPerSecond": 0.02756920569597527,
      "resolution": 64,
      "steps": 4,
      "guidanceScale": 5,
      "textureGeneration": false,
      "driverAllocatedMemoryBytes": 35986374656,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 49582.61379200849,
      "artifactRevision": "f90a0f7df7d5e6f71109cf333f6a95a0ae3194a6",
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "Analytic primitive surface error, finite topology, closed mesh and nondegenerate bounds",
      "score": 0.9999146948718662,
      "threshold": 0.85,
      "meanSurfaceError": 0.00008530512813380528,
      "watertight": true,
      "scope": "Sphere/cube procedural pilot; not arbitrary-object or texture certification"
    },
    "runtime": {
      "engine": "hy3dgen",
      "version": "f8db63096c8282cb27354314d896feba5ba6ff8a",
      "torchVersion": "2.6.0",
      "transformersVersion": "4.51.3",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "model": [
          "mps:0/torch.float32"
        ],
        "vae": [
          "mps:0/torch.float32"
        ],
        "conditioner": [
          "mps:0/torch.float32"
        ]
      },
      "loadMs": 0,
      "accelerationDetails": "Shape DiT, VAE and DINOv2 on selected device; preprocessing, marching cubes and GLB encoding on CPU. No texture generation.",
      "inferenceMs": 35433.30016700202,
      "durationMs": 35483.72925000149,
      "meshesPerSecond": 0.028222039586684348,
      "resolution": 64,
      "steps": 4,
      "guidanceScale": 5,
      "textureGeneration": false,
      "driverAllocatedMemoryBytes": 40505737216,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 35483.72925000149,
      "artifactRevision": "f90a0f7df7d5e6f71109cf333f6a95a0ae3194a6",
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "Analytic primitive surface error, finite topology, closed mesh and nondegenerate bounds",
      "score": 0.99262621481631,
      "threshold": 0.85,
      "meanSurfaceError": 0.0073737851836900535,
      "watertight": true,
      "scope": "Sphere/cube procedural pilot; not arbitrary-object or texture certification"
    },
    "runtime": {
      "engine": "hy3dgen",
      "version": "f8db63096c8282cb27354314d896feba5ba6ff8a",
      "torchVersion": "2.6.0",
      "transformersVersion": "4.51.3",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "placement": {
        "model": [
          "mps:0/torch.float32"
        ],
        "vae": [
          "mps:0/torch.float32"
        ],
        "conditioner": [
          "mps:0/torch.float32"
        ]
      },
      "loadMs": 0,
      "accelerationDetails": "Shape DiT, VAE and DINOv2 on selected device; preprocessing, marching cubes and GLB encoding on CPU. No texture generation.",
      "inferenceMs": 36887.988458009204,
      "durationMs": 36927.39100000472,
      "meshesPerSecond": 0.027109095448192642,
      "resolution": 64,
      "steps": 4,
      "guidanceScale": 5,
      "textureGeneration": false,
      "driverAllocatedMemoryBytes": 40505737216,
      "device": "mps",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 36927.39100000472,
      "artifactRevision": "f90a0f7df7d5e6f71109cf333f6a95a0ae3194a6",
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "Analytic primitive surface error, finite topology, closed mesh and nondegenerate bounds",
      "score": 0.9926267459422439,
      "threshold": 0.85,
      "meanSurfaceError": 0.00737325405775608,
      "watertight": true,
      "scope": "Sphere/cube procedural pilot; not arbitrary-object or texture certification"
    },
    "runtime": {
      "engine": "hy3dgen",
      "version": "f8db63096c8282cb27354314d896feba5ba6ff8a",
      "torchVersion": "2.6.0",
      "transformersVersion": "4.51.3",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "model": [
          "cpu/torch.float32"
        ],
        "vae": [
          "cpu/torch.float32"
        ],
        "conditioner": [
          "cpu/torch.float32"
        ]
      },
      "loadMs": 12702.935290988535,
      "accelerationDetails": "Shape DiT, VAE and DINOv2 on selected device; preprocessing, marching cubes and GLB encoding on CPU. No texture generation.",
      "inferenceMs": 103869.09908402595,
      "durationMs": 103908.66504202131,
      "meshesPerSecond": 0.009627502393094215,
      "resolution": 64,
      "steps": 4,
      "guidanceScale": 5,
      "textureGeneration": false,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 116611.60033300985,
      "artifactRevision": "f90a0f7df7d5e6f71109cf333f6a95a0ae3194a6",
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "Analytic primitive surface error, finite topology, closed mesh and nondegenerate bounds",
      "score": 0.999914694930668,
      "threshold": 0.85,
      "meanSurfaceError": 0.00008530506933206575,
      "watertight": true,
      "scope": "Sphere/cube procedural pilot; not arbitrary-object or texture certification"
    },
    "runtime": {
      "engine": "hy3dgen",
      "version": "f8db63096c8282cb27354314d896feba5ba6ff8a",
      "torchVersion": "2.6.0",
      "transformersVersion": "4.51.3",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "model": [
          "cpu/torch.float32"
        ],
        "vae": [
          "cpu/torch.float32"
        ],
        "conditioner": [
          "cpu/torch.float32"
        ]
      },
      "loadMs": 0,
      "accelerationDetails": "Shape DiT, VAE and DINOv2 on selected device; preprocessing, marching cubes and GLB encoding on CPU. No texture generation.",
      "inferenceMs": 104008.84479098022,
      "durationMs": 104061.48933299119,
      "meshesPerSecond": 0.009614566934279816,
      "resolution": 64,
      "steps": 4,
      "guidanceScale": 5,
      "textureGeneration": false,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 104061.48933299119,
      "artifactRevision": "f90a0f7df7d5e6f71109cf333f6a95a0ae3194a6",
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "Analytic primitive surface error, finite topology, closed mesh and nondegenerate bounds",
      "score": 0.9926267459422439,
      "threshold": 0.85,
      "meanSurfaceError": 0.00737325405775608,
      "watertight": true,
      "scope": "Sphere/cube procedural pilot; not arbitrary-object or texture certification"
    },
    "runtime": {
      "engine": "hy3dgen",
      "version": "f8db63096c8282cb27354314d896feba5ba6ff8a",
      "torchVersion": "2.6.0",
      "transformersVersion": "4.51.3",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "placement": {
        "model": [
          "cpu/torch.float32"
        ],
        "vae": [
          "cpu/torch.float32"
        ],
        "conditioner": [
          "cpu/torch.float32"
        ]
      },
      "loadMs": 0,
      "accelerationDetails": "Shape DiT, VAE and DINOv2 on selected device; preprocessing, marching cubes and GLB encoding on CPU. No texture generation.",
      "inferenceMs": 103620.91233400861,
      "durationMs": 103661.46870900411,
      "meshesPerSecond": 0.009650561623860533,
      "resolution": 64,
      "steps": 4,
      "guidanceScale": 5,
      "textureGeneration": false,
      "driverAllocatedMemoryBytes": null,
      "device": "cpu",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 103661.46870900411,
      "artifactRevision": "f90a0f7df7d5e6f71109cf333f6a95a0ae3194a6",
      "grade": "moderate"
    }
  }
]

Something not quite right?

Tell us what happened with Hunyuan3D-2. 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.