Local AI in a Web Worker
After the model is cached, inference runs in a browser Worker. The image bitmap never goes to ImgSplit’s servers.
AI Background Remover
Remove backgrounds in a local Worker
A one-time on-device model download lets a browser Web Worker cut the subject out. Your photo is not uploaded — processing stays on this computer.
Loading editor...
Features
Private cutouts on-device
One-time model download
The first run fetches the local RMBG model and stores it in this browser’s cache. Later edits usually skip the download.
Edge cleanup controls
Set alpha cutoff, shrink/expand, and feather before you run — then re-remove if you change the sliders.
Transparent export, no upload
Download a transparent PNG or WebP, or a JPEG on white. Batch mode queues several images and zips the ready ones.
How it works
Four short steps
01Drop a photo
PNG, JPEG, or WebP, up to 20MB. Add more files if you want a batch queue.
02Download the model once
If this browser has no cache yet, the first run downloads the local model. The photo still stays on the device.
03Refine the edge
Adjust leftover haze, subject inset, and softness, then run or re-run removal in the Worker.
04Export
Save a transparent PNG/WebP or a white-background JPEG. Ready batch items download as a ZIP.
FAQ
Questions answered
What “local AI in the browser” means for your photos.
No. The image is decoded and processed in this tab’s Web Worker. Only the public model files are downloaded, not your photo.
Background removal needs a local ONNX model (RMBG). It is fetched once, cached in the browser, and reused. Your images are not part of that download.
Transparent PNG, transparent WebP, or JPEG with a white background. Quality applies to WebP and JPEG.
Yes. Add more images to the queue, set edge cleanup once, then start batch removal and download a ZIP of finished cutouts.