1. Quick summary
- Local only: Pixels processed with Canvas; no upload.
- Color detection: Samples corners + edges to guess background.
- Manual picker: Click the preview or use system EyeDropper.
- Threshold + feather: Controls edge precision & softness.
- Batch mode: Queue multiple images and process sequentially.
- Free & no account: Ad supported; instant use.
Try it now: Remove background tool.
2. How the algorithm works
For each image we decode to a Canvas, sample several corner + edge pixels, average them (discarding outliers) to estimate the dominant background color. Each pixel distance to that color (squared RGB delta) is compared to a threshold. Inside threshold → alpha=0 (transparent). Between threshold and threshold+feather → linear alpha ramp to avoid harsh edges.
if (dist2 <= T^2) alpha = 0; else if (dist2 <= (T+F)^2) alpha = ramp;3. Manual color picking & suggestions
You can override auto detection: click on the mini preview to sample a pixel or use the EyeDropper API (in supported browsers). The tool also scans edge strips, quantizes colors and offers up to 6 frequent swatches. Picking a color disables auto detect for precise control.
4. Batch & performance
Processing is O(n) on pixel count. A 2000×2000 image (4M px) is typically processed in tens of milliseconds on modern hardware. Multiple images are processed sequentially so the UI remains responsive.
5. Local privacy & offline
All work is in memory; temporary ObjectURLs are revoked after use. After initial script load you can disconnect the network and still remove backgrounds.
6. Tips for clean edges
- Start with threshold ≈ 35–45 for white backgrounds.
- Increase threshold slowly (steps of 5) until edges clear.
- Add feather (8–18) to reduce blocky transitions.
- For dark backgrounds start slightly lower (25–35) then adjust.
- Avoid overshooting threshold (causes holes in interior light areas).
7. FAQ
- Does it upload my image?
No. All sampling & pixel manipulation happen in your browser memory. - What backgrounds work best?
Uniform (solid or near-same RGB) colors: white, light gray, black, brand swatches. - Why not complex scenes?
The current algorithm is color distance based; multi‑texture backgrounds require advanced segmentation (future AI path). - How to remove halos?
Increase threshold slightly and add 2–4 feather points; too much feather causes semi‑transparent edges. - Transparent vs filled background?
The tool makes selected background pixels alpha=0. You can later place over any color. - Offline usage?
After first load of the script you can disconnect: processing is local.