Background Remover
Cut the background out of a photo without the photo ever leaving your device: the AI models are hosted by this site and run in your browser. Pick your trade-off — Fast finds people, click-to-select cuts out any subject you point at, and Best quality runs the largest on-device model for any subject (a bigger download, and the picker says exactly how big). Refine the edge with one slider, touch up with restore/erase brushes, and download a transparent PNG or re-composite onto a solid color — no signup, no watermark, no credits.
Remove a photo's background — locally
PNG, JPG, or WebP up to 20 MB. Each quality tier's model loads on first use, and every option below shows its real download size — all served from this site, and your photo is never uploaded.
Runs entirely in your browser — verifiably
Most background removers upload your photo to a server, process it there, and meter the results with credits or watermarks. This one does not: it runs open-source segmentation models — Google's MediaPipe selfie segmenter for fast mode, the MagicTouch interactive segmenter for click-to-select (both Apache-2.0), and BiRefNet_lite (MIT, executed by the MIT onnxruntime-web engine) for best quality — inside your browser, from model files this site hosts itself — no signup, no watermark, no credits, and the photo is never uploaded. That is a checkable claim, not a slogan: open DevTools, watch the network panel while you process a photo, and you will see model files arrive from this site and no image bytes leave the page. The model versions, licenses, and provenance (including file hashes) are documented on the methodology page and pinned in the repository.
Where this tool is honestly weaker
Fast mode runs a person segmentation model. It performs best on photos of people — faces, head-and-shoulders shots, full figures — which is the thumbnail and promo use case this site exists for, and it remains people-tuned. For a sneaker, a dog, or any other subject, switch to “Click the subject”: it runs Google's interactive MagicTouch segmentation model, which cuts out whatever you point at — any subject, still entirely in your browser. The two fast modes are weaker on fine details like stray hair than heavyweight server-side services, which run much larger models on much larger hardware; that is the honest trade-off of keeping everything on your device — not a defect this page will pretend away. The refinement pipeline (a hard-edged cut with symmetric feathering, plus restore/erase brushes) usually gives a usable cutout on the first try; whether it does for your photo is something you can see in seconds, for free.
“Best quality — any subject” narrows that gap by running a genuinely larger model on your device: BiRefNet_lite, a dichotomous image segmentation model designed for exactly this kind of subject/background cutout. The trade-off is stated right in the picker, from the real file sizes measured at build time: the tier is a ~86.9 MB on first use download (versus ~11.7 MB on first use for fast mode and ~17.7 MB on first use for click-to-select), and the analysis takes a few seconds of real computation — run in a background thread, so the page never freezes while it works. One more honest number: the model analyzes the photo at 512×512 — the browser-sized build of BiRefNet_lite, chosen because the full-resolution version needs more memory than any browser gives a web page (the exact measurements are pinned in the repository) — and the mask is then scaled back up to your photo's real dimensions and refined there. What does not change is the boundary that matters: it is still 100% local, the photo still never leaves your browser, and no tier of this tool ever will upload it. Larger is not magic — even the best tier can miss wisps of hair — so the same Edge slider and brushes apply to its result too.
Fast mode is the same engine that powers the cutout inside the Thumbnail Maker (the other two modes are extras of this page) — use the maker when you want to place the cutout over gradients, outlined text, and shapes on a real thumbnail canvas.
Frequently asked questions
Is my photo uploaded anywhere?
No. All three models — Google’s open-source MediaPipe selfie segmenter and MagicTouch interactive segmenter (run by @mediapipe/tasks-vision 0.10.35, Apache-2.0) and the BiRefNet_lite model behind “Best quality” (MIT, run by onnxruntime-web 1.29.0, also MIT) — are hosted by this site and executed by JavaScript in your browser — even the model files come from this site, not a third-party CDN. Your photo becomes pixels in this tab and never leaves it. You can verify this in your browser’s DevTools network panel: no image bytes are transmitted.
What kinds of photos does it handle well — and badly?
Fast mode uses a person segmentation model, so it is tuned for photos of people — the thumbnail and promo use case this site is built for. For everything else, switch to “Click the subject” (an interactive model that cuts out whatever you point at) or “Best quality — any subject” (BiRefNet_lite, a much larger model built specifically for high-resolution subject cutouts — and a much larger download because of it: ~86.9 MB on first use). All three run fully locally. Even the best tier can miss fine details like stray hair that heavyweight server-side services running far larger models may catch; the Edge slider and the Restore/Erase brushes are there to fix the edges yourself, in your browser.
Why is “Best quality” such a large download?
Because better masks come from a bigger model, and this site refuses to secretly ship your photo to a server instead. The best tier runs BiRefNet_lite — a dichotomous image segmentation model — entirely in your browser, which means your device has to download the model itself (~86.9 MB on first use, plus the onnxruntime WASM engine, all served by this site). The two fast tiers are ~11.7 MB on first use and ~17.7 MB on first use. Every size shown in the picker is measured from the actual files at build time, so those numbers cannot drift from reality. The trade-off is stated where you choose: quality costs download size and processing time — never your photo.
What files are accepted, and is there a size limit?
PNG, JPG, and WebP images up to 20 MB. Very large photos are scaled down to 4096 pixels on the longest edge before processing (the tool tells you when that happens); your original file is never modified.
Why is the output a transparent PNG?
PNG is the only widely supported web format with real alpha transparency, so a cutout you can layer over any design has to be a PNG. JPG cannot store transparency at all. If you want a flat image instead, pick a solid background color and the tool re-composites the cutout onto it — still a PNG, still built locally on a canvas in your browser.
Can I use this for YouTube thumbnails?
Yes — Fast mode is exactly the same engine that powers the cutout inside our Thumbnail Maker, where you can put the result over gradients, outlined text, and shapes on a real 1280×720 canvas (the Click-to-select and Best-quality engines are extras of this page). Do the cutout here if you just want the PNG, or in the maker if you are building the whole thumbnail. The size rules for the final image are on our YouTube thumbnail size guide.
Use photos you have rights to. The file-acceptance rules, filename handling, and the download-size labels are pure, tested code, and every segmentation engine's model version, hash, and license are pinned in the repository and documented on the methodology page.