Background remover
Remove backgrounds from photos of people, products, and objects.
Drag images here or choose files
PNG, JPG or WEBP, up to 50MB each
AI cutout uses the U²-Net and ISNet models from rembg (danielgatis), running locally in your browser via onnxruntime-web.
A practical guide to Background remover
Background Remover offers subject segmentation for general photos and color-key removal for controlled backgrounds. Tolerance and edge feathering help refine transparent edges before PNG export.
Example: isolate a product photo
Use the quality model for a normal scene, inspect hair and soft shadows against the transparency grid, then export PNG and retain the original for later corrections.
Task-specific steps
- 1
Choose an image and decide between subject segmentation and color-key removal.
- 2
Select a quality mode or sample the background color, then refine tolerance and edge feathering.
- 3
Inspect the transparent edge over the preview background and download a PNG when the cutout is clean.
Options that affect the result
- Segmentation models handle varied scenes; color key works best on an even, known backdrop.
- Tolerance expands the removed color range, while feathering softens the cutout boundary.
Limitations to know first
- Hair, translucent objects, motion blur, and foreground colors similar to the background can produce imperfect masks.
- Transparent output requires PNG; JPEG cannot store an alpha channel.
Questions specific to this task
When should I use color-key background removal?
Use it for a flat studio, green-screen, or solid backdrop with good separation from the subject. A segmentation model is better for ordinary scenes.
Why is there a halo around the subject?
The original background color may be blended into anti-aliased edge pixels. Adjust tolerance and feathering conservatively, then inspect fine edges at full size.