Frequently asked questions
If something here does not answer your question, write to us at service@cutoutpic.com.
Privacy
Is my image uploaded to a server?
No. Your browser downloads a segmentation model and runs it on your device, so the image never leaves it — the free preview has no upload path at all. A paid full-resolution export is planned; if you ever choose it, that will be the only case in which an image is sent to a server.
Is there a background remover that does not upload my photos?
Yes — this one. cutoutpic runs the model inside your browser with WebGPU or WebAssembly. You can confirm it yourself: open your browser devtools, watch the Network tab, and process an image. The only downloads are the page, the inference runtime and the model; no request carries your picture.
Do you store or train on my images?
The free preview cannot: the image never reaches us. The planned paid export will keep images only until a short retention period ends, and will never use them to train models, show them publicly or give them to anyone else.
Can I use the results commercially?
Yes. Your images remain yours, and so do the results. We do not train models on them, display them, or pass them to anyone else.
Using the tool
Do I need an account?
No. The free preview opens and works, with no sign-up and no usage limit.
Is it really free? Is there a watermark?
The in-browser preview is free without limits, and the PNG you download carries no watermark. The only paid feature will be a full-resolution export, which is not open yet.
Does it work offline?
After the model has been cached, the processing itself needs no network. The page and the model still have to be loaded once.
Can I remove the background from several images at once?
Not yet. Images are processed one at a time; after a download you can drop the next one without reloading the page. Batch processing is on the roadmap.
Can I paste an image instead of choosing a file?
Yes. Copy an image or a screenshot and paste it anywhere on the page.
Formats and limits
Which formats are supported?
Input: JPEG, PNG and WebP, up to 12 MB. Output: a PNG with an alpha channel, at most 1280 px on the longer side.
Can I use a HEIC photo from an iPhone?
Not directly — HEIC is not accepted. Export or share the photo as JPEG or PNG first, then drop the converted file in. On an iPhone, Settings → Camera → Formats → Most Compatible makes new photos JPEG.
What resolution is the free download?
The free preview is capped at 1280 px on the longer side. That is enough for profile pictures, listings and web use. Print and full-size work need more; a paid full-resolution export is planned but not open yet.
Why is the downloaded PNG smaller than my original?
Because of the preview cap: a 4000 px photo comes back at 1280 px on its longer side. The aspect ratio is preserved and nothing is compressed — PNG is lossless — but the pixel dimensions are reduced.
Can I get a JPG or a white background instead of transparent?
Not from this tool yet: it outputs a transparent PNG only. To add a white or coloured background, open the PNG in any editor and place it over a filled layer. Background replacement is planned.
Quality
Why is my result rougher than I expected?
Your browser is probably running compatibility mode (WebAssembly) rather than graphics acceleration (WebGPU). The refined model only runs on the WebGPU path. The tier in use is shown next to the page heading.
How well does it handle hair, fur and fine edges?
Fine detail is the hardest case for any segmentation model, and both tiers here tend to merge loose strands into a solid edge. The refined tier (WebGPU) gives a tighter outline without the pale halo of the fast tier. Plain, contrasting backgrounds give the cleanest edges.
Why are there white or coloured edges around the subject?
That fringe is leftover background along the edge pixels — most visible when the original background was bright. A darker or evenly lit background helps; if the fringe matters, a quick pass with an eraser or feather tool in an editor removes it.
It says no subject was found. What now?
That message means the model could not find anything separable — usually because the subject blends into the background. Try an image with more contrast between subject and background. A blank result is never returned in place of a real one.
Why does the first image take longer?
The segmentation model has to be downloaded once. It is cached afterwards, so subsequent images start immediately.
Does it work on a phone, Chromebook or Linux?
It should, wherever the browser supports WebAssembly SIMD — current Chrome, Edge, Firefox and Safari do, on any operating system. We test on desktop Chrome; phones and Chromebooks are not tested yet and are usually slower. Browsers that expose WebGPU get the refined tier automatically.