1Upload the images you want cut out — a clear subject against a distinguishable background gives the model the least to guess at.
2Wait while the segmentation model predicts a per-pixel alpha matte; there is no masking for you to do by hand.
3Check the edges in the preview, particularly hair, fur, glass and motion blur, which are where any automatic matte struggles.
4Download the result as PNG or WebP so the transparency actually survives — saving a cut-out as JPG puts the white back.
Remove Image Background FAQ
How does Remove Image Background decide what is background?
+
Concretely, a U^2-Net segmentation model predicts the alpha matte on whatever you upload, and the output is always written to a format that can actually carry transparency rather than silently flattening onto white. There is no manual masking step — a neural segmentation model does the separation and you get a transparent result back.
How accurate is the cut-out on difficult edges?
+
Solid subjects against a distinguishable background come out excellent. Fine hair, fur, motion blur, glass and chain-link fences are where any automatic matte struggles, and where you may still want to touch up the result by hand afterwards.
Does my source format affect the quality of the matte?
+
Yes. this format is read and written by the same streaming pipeline as everything else we support. Compression artefacts around a subject are exactly what a segmentation model has to disambiguate, so cleaner input gives a cleaner edge.
Anything specific to Image I should consider?
+
Yes — the pipeline reads the real format from the file header rather than the extension, so a mislabelled upload still works. That shapes both what the model sees and what the output can carry.
What format should the result be saved in?
+
One with a real alpha channel — PNG or WebP. Saving a cut-out as JPG throws the transparency away and composites the subject onto white, which defeats the entire operation and is the single most common thing that goes wrong here.
Can I process a whole batch of images?
+
Yes. Every image is segmented independently, so a set of product shots can be cut out in one pass — which is normally the point, since doing it by hand is the slow part of a catalogue shoot.
Which image formats does Remove Image Background accept?
+
All of the common ones — JPG, JPEG, PNG, WebP, GIF, BMP, TIFF, SVG, ICO, PSD, JFIF and HEIC. You can mix formats freely in one batch; each file is identified from its header and routed to the right decoder.
Is there a file size limit on Remove Image Background?
+
Yes: free accounts process images up to 25 MB each regardless of which image format you brought; ImageMagick, libjpeg-turbo, libpng and libwebp do the actual work. The cap is per file, so a batch of twenty 5 MB photos is fine even though the total is well over the limit.
Will Remove Image Background lower the quality of my images?
+
That depends on what you brought. A lossless source (PNG, BMP, TIFF) round-trips exactly; a lossy one (JPG, WebP, GIF) is already the product of one quantisation pass and only re-encodes if the operation genuinely requires it.
Can I run Remove Image Background on several images at once?
+
Yes, and mixing formats in the batch is fine — settings are applied per file with the right decoder for each. Free accounts run one file at a time; a pass removes that.
What happens to metadata and transparency?
+
EXIF and the ICC colour profile are carried across wherever the target format supports them, and an alpha channel survives into any format that has one. Transparency is only flattened when the destination format has no way to represent it.
Does Remove Image Background cost anything?
+
No. Remove Image Background is free without an account and adds no watermark. Uploads are deleted from the workers shortly after the job completes.
