How to use
- Choose a model (Portrait, or Objects for products and animals on WebGPU) and press Download model; it is cached for next time.
- Add images and press Remove background.
- Pick a background (transparent, colour, blur or image), adjust edge position and feather, compare before and after, and download.
Worked example
The 800×1000 sample portrait with the Portrait model becomes a transparent PNG: the person (face, hair and shirt) stays opaque and the corners and wall panels become fully transparent.
Supported formats and limits
| Input | JPEG, PNG, WebP |
|---|---|
| Output | PNG (transparent), PNG or JPEG on a new background |
| Limits | First use downloads a model: about 12 MB (Portrait) or 109 MB (Objects, needs WebGPU). Up to 40 megapixels and 50 images per batch. |
| Engine | transformers.js with MODNet (Apache-2.0, fp16) or BiRefNet-lite (MIT, fp16) in a Web Worker; WebGPU when available, otherwise WebAssembly (Portrait only) |
Limitations
- Fine hair, glass and very low-contrast edges may need manual cleanup.
- The Portrait model is trained on people; use Objects (WebGPU only) for products, animals and things.
- Without WebGPU the Portrait model runs on the CPU: a few seconds per photo on a typical laptop.
Questions
Which model should I choose?
Portrait (about 12 MB) is trained on people and works with or without WebGPU. Objects (about 109 MB) handles products, animals and things, and needs a browser with WebGPU.
What should I check before downloading?
Look at the edges with the before and after view. Fine hair, glass and low-contrast edges can be wrong; the edge position and feather sliders help, but some images still need manual cleanup elsewhere.
Privacy
On-device model. Processing runs in this browser. The open-source model files are downloaded once from the model host (Hugging Face) and cached; your content is not uploaded.
- Hugging Face: Only after you press Download model: requests for the model files (MODNet about 12 MB, or BiRefNet-lite about 109 MB) from huggingface.co and its CDN, which see your IP address. Your images are never sent.
See the privacy policy for how toolsdocks handles data.