Blur the background without upload
Portrait mode as a retrofit: the Depth Anything AI estimates how far away every point in the photo is and softens everything that is not in the focus plane – the farther away, the stronger, the way a fast lens renders it. Tap what should stay sharp and set the blur and the depth of field. Everything happens in your browser, with no upload and no third parties.
Tap the part of the image that should stay sharp. With every new photo the focus starts on the subject in the middle.
Your browser does not support WebGL – the 3D view is not available.
AI model Depth Anything V2 · one-time ≈ 25 MB from mytoolio.org
No photo yet. Drop an image here – the AI estimates how far away every point is and softens everything that is not in the focus plane.
How it works
Choose a photo
Drop a photo into the box, pick it or paste it with Ctrl+V. The AI estimates the depth – in about a second with a graphics card.
Set the focus
The focus starts on the subject in the middle. Tap the image to move it, and set how strongly the rest blurs and how deep the sharp zone reaches.
Compare & save
Hold “Original” to compare before and after, and save the picture at full resolution – as JPG; PNG stays PNG.
Frequently asked questions
Is my photo uploaded?+
No. Your photo is only read in your browser and analysed there by the AI – it never leaves your device. The AI model (Depth Anything V2) and the runtime come from mytoolio.org itself; there are no requests to other services.
How is this different from an ordinary blur?+
An ordinary blur makes everything behind a cut-out subject equally soft. Here the estimated distance sets the blur: things just behind the subject stay almost sharp, distant things blur strongly, the ground fades gently out of focus – the way a lens with a wide aperture renders it. If you would rather have a cut-out subject in front of an evenly blurred background, use “Remove background” with the “Blurred” background.
How do I decide what stays sharp?+
Tap the spot in the image that should be sharp – or move the “Focus plane” slider between back and front. “Depth of field” sets how much in front of and behind it stays sharp: narrow for a single face, wide for a group in several rows. Everything closer to the camera than the focus plane stays sharp; with “Blur the foreground as well” it blurs too.
How clean are the edges?+
Very clean for clear subjects in front of a calm background: the depth map is fitted to the edges in the photo. Very fine details such as single hairs, fences or the arms of glasses may get a little blurred. “Depth detail: Fine” picks up smaller structures but takes a good twice as long. Mirrors, panes of glass and pictures within the picture can fool the AI.
Does it only work with portraits?+
No. It works with anything that has spatial depth: people, animals, products on a table, flowers, cars, food. With flat subjects such as documents, screenshots or landscapes without a foreground there is hardly any difference in depth – the effect is weak there.
Is the resolution kept, and how long does it take?+
Yes, the picture is saved at full resolution (up to 6,000 pixels on the long side, 4,096 on phones); metadata such as the location is not included in the new file. With WebGPU (current Chrome or Edge) the depth estimate takes about a second, otherwise the processor needs 10 to 20 seconds. After that the blur responds instantly to every slider.
Depth estimation with Depth Anything V2 Small (Apache 2.0, weights stored as int8) via ONNX Runtime Web (MIT) – both served by mytoolio.org and run entirely in your browser. The depth is an estimate; at fine edges the blur can be imprecise.