How AI can remove background from a photo
The tool uses a model based on U2Net, a neural network trained to spot the main subject in a picture. It looks at every pixel and decides whether it belongs to the subject or to the scene behind it. The result is a mask with soft edges, which is then used to cut the subject out.
You get the cleanest cutout when:
- there is one clear subject that fills a good part of the frame
- the subject and the background have different colors
- the light is even and the subject is in focus
Flyaway hair, glass, smoke and sheer fabric are hard for any automatic tool. If an edge looks rough, try another shot with a plainer background. There is no manual brush for touch ups yet, so the source photo matters most.
Choose the right output
| Option | Saved as | Good for |
|---|---|---|
| Transparent | PNG | Designs, stickers, slides, logos on colored pages |
| White | JPG | Product listings, catalogs, resumes, team pages |
| Black or Color | JPG | Social posts in brand colors, bold product shots |
| Blur | JPG | Profile pictures that still look like the real place |
If you are not sure yet, take the transparent PNG. You can place it on any background later, in PowerPoint, Keynote or your shop editor.
Each background has its own page with more tips: White Background for shop photos and Blur Background for portraits. A transparent PNG of a large photo can weigh several megabytes, so run it through Compress PNG before you put it on a website.
Where people use cutouts
- Online shops: a Shopify store owner shoots ten mugs on a kitchen table and puts all of them on the same white.
- Presentations: a cutout of a phone or a team member sits on a slide without a gray box around it.
- Thumbnails: a creator places themselves in front of bold text for a video cover.
- Cards and collages: a family photo, the dog from another picture and a birthday banner end up in one image.