How it works
The image is scaled to 320×320, a small U²-Net model predicts the subject mask, and the mask is scaled back to the original size and used as the alpha channel. The result is a transparent PNG; the original file is not changed.
When to use it
Product photos for a shop, profile pictures, stickers and thumbnails, or a clean cut-out for a slide. It works best when the subject is clearly separated from the background.
Tips
- Use a well-lit photo with a clear edge; the model can struggle with hair, glass and motion blur.
- The first run downloads the model and runtime (about 28 MB) once; later runs reuse them from the browser cache.
- For a shop, turn on the white background if the marketplace does not accept transparency.
- Everything runs on your device. Large batches take longer on a phone.
Frequently asked questions
Is my image uploaded?
No. The image and the AI model run entirely in your browser. Nothing is sent to beancuit or anyone else.
Which images work best?
Clear subjects with a distinct edge, such as products, people, pets and logos. Very fine hair, glass and busy backgrounds can leave rough edges.
Can I process several images at once?
Yes. Add up to 20 images and download them one by one or as a ZIP. The first run downloads the AI model and runtime (about 28 MB).
Is it really free?
Yes. The model and runtime are open source, and the tool runs on your device, so there is no per-image fee.