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.

ORMBG — Human Background Removal

Foreground cutouts from images. Processed on your device.

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

Separate people from an image background for background removal or replacement. Returns a foreground matte and transparent-image output; designed primarily for human subjects.

License
Apache-2.0
Model released
Unknown
Source repository created
May 23, 2024
Model weights · FP16
88.1 MB

Best for

  • Removing or replacing the background of a single-person portrait.
  • Creating a soft alpha matte for compositing portraits in browser applications.
  • Prototyping client-side portrait segmentation without a hosted inference API.

Keep in mind

  • ORMBG is specialized for images containing humans and is not a general object-segmentation model.
  • The upstream author reports that the model can struggle on real images because synthetic training data does not fully represent real lighting, occlusion, scale, and backgrounds.
  • Hair, transparent materials, motion blur, occlusion, multiple people, unusual poses, and low foreground/background contrast can produce incomplete or haloed masks.
  • The output is an estimated matte rather than verified ground truth and must not be used for identity, biometric, medical, or safety-critical decisions.
  • Higher processing resolutions require more memory and computation and can fail on constrained devices.

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
schirrmacher/ormbg
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
Sep 29, 2024
Browser model files
onnx-community/ormbg-ONNX
Files updated
Apr 14, 2026
Artifact revision
034e2d884afbab897e10e78fc5bb566b29533fd6

Input

PNG / JPG / WEBP

Sample result

ORMBG — Human Background Removal

Settings

Adds a self-contained SVG presentation image alongside the unchanged native model fields. When enabled, draws model predictions or presents generated media and text. Adds CPU postprocessing time and image bytes. It does not change model predictions.
Advanced settings 2
Image processing size; use a multiple of 32. Lower sizes reduce work and memory but can lose edge detail. Higher sizes can retain more detail at greater cost; dimensions must be multiples of 32.
Automatic preserves soft edges. Set a value for a hard mask. Automatic keeps soft alpha edges. Lower cutoffs keep more foreground pixels; higher cutoffs remove faint edges and can lose fine detail.

Output

Your cutout will appear here

Choose an image or try the sample, then remove its background. Compare the original and download a transparent PNG.

Benchmark evidence

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

ORMBG — Human Background Removal

16 configurations tested
Passed16
Failed0
TestedSep 8, 2026
RuntimeTransformers.js 4.2.0

Quality criteria

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

PrecisionDeviceResult
FP16GPU · WebGPUAcceleratedPassedFast
Q4F16GPU · WebGPUAcceleratedPassedFast
Q4GPU · WebGPUAcceleratedPassedFast
BNB4GPU · WebGPUAcceleratedPassedFast
FP32GPU · WebGPUAcceleratedPassedFast
UINT8GPU · WebGPUAcceleratedPassedVery slow
Q8GPU · WebGPUAcceleratedPassedVery slow
INT8GPU · WebGPUAcceleratedPassedVery slow
FP32CPU · WASMPassedVery slow
Q4CPU · WASMPassedVery slow
BNB4CPU · WASMPassedVery slow
Q4F16CPU · WASMPassedVery slow
FP16CPU · WASMPassedVery slow
UINT8CPU · WASMPassedVery slow
Q8CPU · WASMPassedVery slow
INT8CPU · WASMPassedVery slow
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-09-08T17:07:37.483Z
Environment
Headless Chrome 151 on macOS; Transformers.js 4.2.0
Runtime
Transformers.js 4.2.0
Artifact revision
034e2d884afbab897e10e78fc5bb566b29533fd6
Fixture
{
  "file": "modnet-portrait-pexels-5965592.jpg",
  "url": "https://f005.backblazeb2.com/file/hubters-webai-public/assets/benchmarks/modnet-portrait-pexels-5965592.jpg",
  "source": "https://www.pexels.com/photo/woman-wearing-pink-turtleneck-sweater-5965592/",
  "license": "Pexels License",
  "size": 78124,
  "sha1": "38212739a4a4f0be25d84863cf991d7fb6c6d818",
  "width": 1024,
  "height": 683
}
Rerun when
model artifact revision change; Transformers.js or ONNX Runtime version change; benchmark fixture, reference matte, or quality-gate change; target browser/backend behavior change
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0,
      "intersectionOverUnion": 1,
      "foregroundRatio": 0.256265,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 38114.8,
      "inferenceMs": 7085.2,
      "totalMs": 45230.9,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0987,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0,
      "intersectionOverUnion": 1,
      "foregroundRatio": 0.256265,
      "dynamicRange": 0.996078
    },
    "runtime": {
      "loadMs": 24680.2,
      "inferenceMs": 264.1,
      "totalMs": 24975.5,
      "grade": "slow",
      "megapixelsPerSecond": 2.6482,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000002,
      "intersectionOverUnion": 1,
      "foregroundRatio": 0.256265,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 19426.7,
      "inferenceMs": 7291.1,
      "totalMs": 26744.6,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0959,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000887,
      "intersectionOverUnion": 0.999939,
      "foregroundRatio": 0.256267,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 18024,
      "inferenceMs": 224.2,
      "totalMs": 18278.2,
      "grade": "good",
      "megapixelsPerSecond": 3.1195,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "int8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000274,
      "intersectionOverUnion": 0.999286,
      "foregroundRatio": 0.256191,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 21876.1,
      "inferenceMs": 23701.3,
      "totalMs": 45606.6,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0295,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "int8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.00028,
      "intersectionOverUnion": 0.999297,
      "foregroundRatio": 0.256191,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 10103.7,
      "inferenceMs": 24517.2,
      "totalMs": 34648.2,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0285,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "q8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000208,
      "intersectionOverUnion": 0.999459,
      "foregroundRatio": 0.256336,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 8124.6,
      "inferenceMs": 7564.2,
      "totalMs": 15716.3,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0925,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000212,
      "intersectionOverUnion": 0.999442,
      "foregroundRatio": 0.256337,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 9839.2,
      "inferenceMs": 7838.9,
      "totalMs": 17704.8,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0892,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000208,
      "intersectionOverUnion": 0.999459,
      "foregroundRatio": 0.256336,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 8970.6,
      "inferenceMs": 7483.1,
      "totalMs": 16482.2,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0935,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000212,
      "intersectionOverUnion": 0.999442,
      "foregroundRatio": 0.256337,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 9247.6,
      "inferenceMs": 7817.3,
      "totalMs": 17091.5,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0895,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0,
      "intersectionOverUnion": 1,
      "foregroundRatio": 0.256265,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 24528.2,
      "inferenceMs": 7138.1,
      "totalMs": 31696.1,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.098,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0,
      "intersectionOverUnion": 1,
      "foregroundRatio": 0.256265,
      "dynamicRange": 0.996078
    },
    "runtime": {
      "loadMs": 23564.2,
      "inferenceMs": 255.4,
      "totalMs": 23852.2,
      "grade": "slow",
      "megapixelsPerSecond": 2.7384,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000002,
      "intersectionOverUnion": 1,
      "foregroundRatio": 0.256265,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 15641.7,
      "inferenceMs": 7235,
      "totalMs": 22907.6,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0967,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0.000887,
      "intersectionOverUnion": 0.999939,
      "foregroundRatio": 0.256267,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 22414.4,
      "inferenceMs": 229.4,
      "totalMs": 22673.8,
      "grade": "good",
      "megapixelsPerSecond": 3.0488,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "bnb4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0,
      "intersectionOverUnion": 1,
      "foregroundRatio": 0.256265,
      "dynamicRange": 1
    },
    "runtime": {
      "loadMs": 24034.9,
      "inferenceMs": 7169.3,
      "totalMs": 31233.2,
      "grade": "very-slow",
      "megapixelsPerSecond": 0.0976,
      "loadGrade": "slow"
    }
  },
  {
    "precision": "bnb4",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "grade": "excellent",
      "meanAbsoluteError": 0,
      "intersectionOverUnion": 1,
      "foregroundRatio": 0.256265,
      "dynamicRange": 0.996078
    },
    "runtime": {
      "loadMs": 24596.2,
      "inferenceMs": 257.3,
      "totalMs": 24883.6,
      "grade": "slow",
      "megapixelsPerSecond": 2.7182,
      "loadGrade": "slow"
    }
  }
]

Something not quite right?

Tell us what happened with ORMBG. 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.