Background remover
On-device AI removes the background from photos of people — the image is processed entirely in your browser and never uploaded.
Drop a photo here
or click to browse · works best with people and portraits
The AI runs on your device — your photo is never uploaded.
How it works
- Drop a photo into the box. The AI model (about 3 MB) loads once, directly in your browser.
- The background is removed automatically. Fine-tune the edge with the threshold and feather sliders, or switch the background to a solid color, a blur, or your own image.
- Download the result as a transparent PNG, or as JPG when using a solid or blurred background.
Frequently asked questions
- Is my photo uploaded to a server?
- No. Unlike most background removal sites, the AI model runs inside your browser using WebAssembly. The photo, the mask and the result never leave your device — which also means there are no upload limits, watermarks or paid credits.
- What kinds of photos work best?
- The on-device model is trained for people — portraits, selfies and group shots give the cleanest results. For product photos or logos on a plain light background, the Remove White Background tool usually gives a sharper cutout.
- Why are the edges rough on my photo?
- Increase the feather slider to soften the transition, or adjust the threshold to include or exclude borderline pixels. Very busy backgrounds, low light and hair are the hardest cases for any segmentation model, including the large server-based ones.
- What resolution is the output?
- The mask is computed by the model and then applied to your photo at its full original resolution, so the downloaded PNG keeps every pixel of the source image. Nothing is downscaled.
- Is this really free? What’s the catch?
- It is free without a catch: because processing happens on your device, it costs us nothing per image, so there are no credits, sign-ups or watermarks. The site is supported by unobtrusive ads instead.
About this tool
This background remover runs a real segmentation neural network — Google’s MediaPipe selfie segmentation model — directly in your browser through WebAssembly. When you drop a photo, the model estimates for every pixel how likely it is to belong to a person, and that probability map becomes the transparency mask of your image. Because everything happens on your device, there is nothing to upload, no queue to wait in, and no server that ever sees your picture.
The result is editable rather than take-it-or-leave-it. A threshold slider decides how confident the model must be before a pixel is kept, and a feather slider softens the boundary so hair and shoulders blend naturally instead of looking cut out with scissors. The background itself can be fully transparent for use in designs, a solid color for ID-style photos, a blurred version of the original for a portrait-mode look, or any image you choose.
Typical uses: profile pictures with a clean backdrop, product-style headshots, stickers and memes, e-commerce sellers photographing themselves wearing items, and slide decks that need people cut out of snapshots. For the common case of a logo or scan on a plain white background, the dedicated Remove White Background tool on this site is faster and pixel-exact.
Honest limits: the on-device model is optimized for people, so pets, cars and products may segment poorly; extremely cluttered scenes and single strands of hair challenge every background remover. The model file is about 3 MB and loads once per session — after that, each photo processes in roughly a second on a modern device.