Portrait matting
MODNet predicts a soft alpha matte around people, preserving flyaway hair far better than a hard threshold.
AI retouch · CutoutsLocal-first
AI Background Removal
Portrait matting with MODNet and general subject selection with OpenCV GrabCut — both local.
Portraits and products need different cutout tools. MODNet is a portrait matting network that preserves hair and soft edges, while OpenCV GrabCut is a general-purpose foreground selector for objects, products and scenes.
Both paths run locally. The result becomes a mask on the active layer, so it can be refined by hand, combined with a new background, or exported as a transparent PNG.
No image uploads WebAssembly engine Works in the tab
What you can do
Background removal, in detail.
Remove backgrounds locally in your browser. MODNet makes soft portrait mattes; GrabCut selects general subjects. Export a transparent PNG without uploading.
General subject selection
GrabCut uses a foreground/background colour model to isolate products, animals and other non-portrait subjects.
Editable masks
The cutout is not baked into the pixels. Refine it with the brush, feather the edge, or combine it with other masks.
Transparent export
Export a PNG with alpha for compositing, or place a new background layer underneath in the same document.
Local and private
The segmentation models run in the tab through ONNX Runtime WebAssembly. No cutout is uploaded.
Works with the layer stack
Use the cutout as a mask for colour grading, a shadow, a stroke or a completely new backdrop.
How it works
From file to finish in three moves.
No install step and no import queue. Open the editor, work on the document locally, and export the file your next tool expects.
Try it now- 1
Choose the cutout mode
Use Portrait for people and Subject for everything else. Draw a rough selection if GrabCut needs a hint.
- 2
Run the model
The image is downscaled for inference, then the alpha mask is scaled back to document resolution.
- 3
Refine and export
Clean up the mask with the brush and export a transparent PNG or composite a new background.
Who it is for
Built for real photo work.
These are the workflows the tool is designed around — not a list of theoretical features.
- Wedding and portrait photographers creating clean cutouts for albums and invitations.
- Product photographers replacing a studio backdrop without a cloud service.
- Social media editors producing transparent subject stickers.
- Designers who need a quick alpha matte before moving into a layout tool.
Questions
Background removal FAQ
Straight answers about formats, privacy and limits.
Which model should I use for people?
Portrait mode uses MODNet and generally gives the best hair and edge detail on people. Subject mode with GrabCut is better for products and non-human subjects.
Can I remove the background from a group photo?
MODNet produces a matte for the salient person or people in the frame. Complex groups with overlapping subjects may need manual mask cleanup.
Does it work in batch?
Background removal is an interactive document operation in the current build. The batch processor focuses on decode, tone, resize, sharpen and encode; it does not run matting on every file.
Is the cutout uploaded?
No. Inference runs locally. The only download is the model weights, which are cached after the first use.
Put background removal to work.
Open the editor and bring a real file. Your photo stays on your device.