How to use Background Remover
- Choose one still JPG, PNG or WebP up to 10 MiB, 16 megapixels and 8,192 pixels per side, or try the bundled artwork.
- Select Transparent or Solid color for the output background. Choose white for a white background, or pick another color.
- Select Remove background. The first run downloads a 4.6 MB model and an 11.2 MB runtime from this website, then processes your image locally. Cancel stops the worker.
- Use the comparison slider or side-by-side view to inspect the original and cutout. Check the subject, fine edges and any areas the model removed incorrectly.
- Download the PNG at the original pixel dimensions. Changing the output background creates an updated PNG without running segmentation again.
Questions about this tool
How do I make an image background transparent?
Choose your image, leave Output background set to Transparent, then select Remove background. The tool runs the U2Netp segmentation model locally and uses its foreground mask as PNG transparency. Inspect the cutout before downloading; an automatic mask can remove parts of the subject or retain parts of the background.
Can I give a product photo a white background?
Yes. Select Solid color and choose white (#FFFFFF), then remove the background. The downloadable PNG contains the cutout over that color. You can change the color after removal; this recomposes the existing cutout without estimating a new mask. The original image is not modified.
Are my photos uploaded to a server?
No. This tool decodes your selected image, runs the model and exports the PNG inside a browser worker on your device. It does not send image contents or filenames to an upload API. The model, runtime and optional sample are downloaded from this website. The browser may cache those public assets; the tool does not save your selected image to browser storage.
Why does the first removal download files?
Local AI processing needs model weights and a WebAssembly runtime. The model is 4,574,861 bytes, about 4.6 MB, and the runtime binary is 11,210,254 bytes, about 11.2 MB; JavaScript adds a small amount. Downloads come from this website and may be reused by the browser cache. A successful download is followed by local processing, whose duration depends on the device. Offline availability is not guaranteed.
Which photos work best?
Start with a clearly visible foreground object that contrasts with the background. This compact model estimates a 320 × 320 mask and scales it to the original dimensions. It may lose fine hair, fur, lace, transparent glass, soft shadows or low-contrast edges, and it may select an unintended object. The tool has no brush or manual mask editor; use the comparison before relying on the result.
Will the PNG keep my original image dimensions?
Yes, within the input limits. The output uses the decoded original width and height; the subject is not cropped or enlarged. The foreground mask is estimated at lower resolution, so preserving output dimensions does not restore missing edge detail. Existing transparent pixels remain transparent in transparent mode. Solid-color mode fills all transparent areas.
Which file types and sizes are supported?
Input supports one still JPG, PNG or WebP up to 10 MiB, 16 million pixels and 8,192 pixels per side. Header dimensions are checked before full decoding. Animated PNG and WebP, GIF, SVG and HEIC are not accepted. Output is PNG, with a 64 MiB maximum. Browser memory limits can still prevent a large image from processing; try a smaller image if that happens.
Why is the download PNG instead of JPG?
PNG can store transparent pixels and soft alpha edges. JPG cannot retain transparency. This tool always exports PNG, including when you choose a solid background, so the comparison and download use one consistent format. It does not add a watermark or require an account.
What happens when I cancel or clear an image?
Cancel terminates the active worker and ignores late results. Your original image stays available for another attempt. Clear or leaving the page releases image URLs, outputs and the worker, including its in-memory model. Public model and runtime files may remain in the browser cache, and files you already downloaded remain on your device.
Does the output retain EXIF, location data or color profiles?
The browser applies supported EXIF orientation while decoding. Export redraws the pixels into a new PNG and does not copy the original EXIF or GPS metadata. Color handling depends on the browser; this tool is not intended for color proofing. File size may increase even when the background is removed.