Blur the background of any photo automatically and get a portrait-mode look with the subject kept sharp. An AI model finds the person or object in your browser, then the background is blurred by the amount you choose. Process a batch of photos and download them all — with no upload, no account and no watermark.
| Setting | Look |
|---|---|
| 0.5% | Gentle, like f/4 |
| 1.5% | Clear portrait effect |
| 3% | Strong bokeh |
| 5% | Background unrecognizable |
Drop the photo, let the model find the subject, and choose how strong the blur should be. Then download.
No. The AI model runs in your browser; only the model itself is downloaded, once.
Yes. Moving the slider updates the result instantly without re-running the AI.
Yes. Choose a general or object model for products and pets.
Fine hair is hard to separate. The feather and tightness controls help; blur hides most edge errors anyway.
A segmentation model runs in the browser through transformers.js (WebGPU where available, otherwise WebAssembly) to find the subject and produce a soft alpha matte at full photo resolution. The output is composited in two layers: first the whole photo, made fully opaque, is drawn with a Gaussian blur whose radius is a percentage of the photo's long side — slightly enlarged so the blurred edges don't fade to transparent — and then the cut-out subject is drawn on top using the matte. The matte's soft edges blend the sharp subject smoothly into the blur, and the edge-feather and tightness controls adjust that transition. Changing the blur strength only re-composites the stored matte, so it's instant after the first run. The model is downloaded once and cached by the browser; photos are never uploaded. Results are saved as high-quality JPG.