Blur Photo Background
Add a DSLR-style bokeh background to any portrait. The subject is detected automatically, the background is blurred, and the result is composited — all inside your browser, with no upload.
Drop a portrait photo here or click to choose a file
JPG, PNG or WebP · the model runs locally after a one-time download
Choose a portrait photo to begin.
How the background blur works
The page loads a segmentation model that runs entirely in your browser via WebAssembly. It returns the subject as a transparent cut-out; the background layer is then blurred and the sharp subject is composited back on top — the same two-layer trick a portrait-mode camera uses.
🔒 Nothing is uploaded
Image decoding, segmentation and compositing all happen on your device. The only network request is the one-time model file.
🎚 Adjustable strength
Slide from 0 px (barely visible separation) to 40 px of heavy bokeh until the subject pops.
🖼 Full resolution
The result keeps the original pixel dimensions, so it stays usable for print and for social platforms.
🧠 Real model, stated honestly
This page uses @imgly/background-removal v1.7.0 (ISNet) for segmentation. Blur itself is a standard canvas operation.
How to blur the background of a photo
- Add a portrait. Photos with a clearly separated person work best; a busy scene with many small objects is harder for any segmentation model to split cleanly.
- Wait for the model on first use. The about 30 MB model is downloaded once and cached by the browser, so later photos process almost instantly.
- Set the blur strength. Start around 12–16 px; push higher for a stronger faux-bokeh look or lower to keep some background context.
- Download the PNG. PNG keeps the full quality of the composite; convert to JPG afterwards if you need a smaller file.
Frequently Asked Questions
Is the blur real or a fake effect?
It is a real composite: the subject is cut out with a neural network, the background is genuinely blurred with a Gaussian filter, and the two layers are merged. Nothing is faked with an overlay.
Why does the subject sometimes get clipped?
Segmentation is a best-effort prediction. Unusual poses, very low contrast between subject and background, or hair against a similar colour can produce a rough edge. Try a photo with more separation.
Does it work on group photos or pets?
Yes — the model segments the salient foreground, which can be several people, a pet, or an object. Results vary with how busy the scene is.
Why is the first run slow?
The browser downloads the model file (about 30 MB) once. Every subsequent photo uses the cached copy, so only the first image waits.
Can I undo the blur?
Yes — this page never modifies your original file. Only the newly generated PNG is saved.