Blur Photo Background
Give any photo a portrait-mode look: the subject stays sharp while the background softens.
- Free
- No upload
- No watermark
- No sign-up
- AI runs on your device
How to blur Photo Background
- 1
Choose a photo of a person, pet or object.
- 2
Press Remove background to find the subject.
- 3
Set how strong the blur should be.
- 4
Download the photo as JPG or PNG.
Portrait mode after the fact
Phone portrait mode uses depth sensors or two lenses to blur what is behind you. This tool gets a similar look from any photo, even an old one or a screenshot: the AI finds the subject, the rest of the picture gets a smooth Gaussian blur, and the sharp subject is laid back on top. Slide the blur from a gentle softening to a strong, creamy background; it redraws instantly.
Unlike a real lens, the blur is the same everywhere in the background rather than growing with distance, so it looks most natural when the subject stands well in front of the background. Edges of hair may show a slight halo on strong settings.
How it works, and what stays private
A segmentation model looks at the photo and decides, pixel by pixel, how much of each belongs to the subject. That soft mask becomes the new transparency, so hair and fuzzy edges fade out instead of being cut with scissors. The model runs inside your browser: the photo is never uploaded, which matters for ID photos, family pictures and unreleased products.
There are two models. For people and portraits it uses ORMBG (Apache-2.0, a 44 MB download), trained on photos of humans. For products, pets and objects it uses BiRefNet lite (MIT, a 99 MB download), a general model that finds any main subject. Your browser keeps the model after the first download. The people model studies the photo at 1024 × 1024 and the general one at 512 × 512; either way the mask is scaled back to your photo's full size, so the result keeps every pixel of the original.
Speed and limits, honestly
- With WebGPU (Chrome or Edge on most laptops) a photo takes a second or two. On a processor the people model takes a few seconds and the general model 10 to 30 seconds.
- It keeps one main subject. Group shots usually work; a busy scene where the subject is not clear may keep or drop the wrong things.
- Fine hair against a similar-colored background, glass and smoke are the hardest cases for every remover, this one included.
- The result is the size of your photo. Very large photos (over about 40 megapixels) may run out of memory on phones.
Questions, answered.
Does it work on old photos without depth data?
Yes. It needs no depth information: the AI finds the subject from the picture alone.
Can I blur a busy background behind a product?
Yes. Choose products and objects as the subject so the general model is used.
Is it really free, with full-resolution downloads?
Yes. The model runs on your device, so there are no credits, no watermark and no low-resolution preview: the download is the full size of your photo.
Is my photo uploaded?
No. The photo is processed inside your browser tab. The only download is the model itself, the first time.
Which formats can I use?
JPG, PNG, WebP, AVIF and BMP in. Out: PNG with transparency, or PNG or JPG with a solid or blurred background.
Why does it take a while the first time?
Your browser downloads the model once (44 MB for people, 99 MB for the general model) and keeps it. After that, photos start right away.
More free image tools
- Remove backgroundCut out people, products and pets in one click, with AI that runs on your own device.
- White background for photosPut a clean white, blue or any color behind your photo for forms, IDs and product listings.
- AI image upscalerMake small images bigger and sharper with Swin2SR, running on your own device.
- HEIC to JPGTurn HEIC images into JPG in seconds, one or hundreds at a time, without uploading them anywhere.
- HEIC to PNGTurn HEIC images into PNG in seconds, one or hundreds at a time, without uploading them anywhere.
- WEBP to JPGTurn WEBP images into JPG in seconds, one or hundreds at a time, without uploading them anywhere.