Remove a photo background with rembg
Run rembg on your computer, save a transparent PNG and use it in a design. Follow the real workflow from the GitHub link to the finished file.
rembg removes the background from a photo and saves the subject as a transparent PNG. UGT runs it on your computer, checks the output and opens the folder where the file was saved.
The walkthrough above starts with rembg’s GitHub link and ends with the actual PNG used in a poster. UGT is already installed on the Mac in the recording. If this is your first project, set up UGT with your AI app, then return here.
Open rembg
Copy https://github.com/danielgatis/rembg into UGT’s GitHub link field. Choose Open in UGT if you have it installed. Your browser may ask you to confirm opening the app.
Prefer help from your AI app? Choose Set up with my AI app. The message carries the selected project and task so you can continue there.
Choose your photo and run it
Select a photo with a clear foreground subject, then choose Run. The project runs locally. The page shows what UGT is doing while it prepares rembg, processes the photo and checks the result.
The first run may need to download the project and its model. Keep the page open, or reopen the same job to check its progress. The video shortens a preparation wait and marks the processing section at 6× speed; the time depends on your setup, photo and computer.
Open the transparent PNG
After the check passes, use Show in Finder on a Mac, or the result-folder action on your computer. The output is cutout.png. Open it to inspect the edge of the subject before using it in another image.
In the film, the white tiger cutout is placed into a separate poster. That design is an example of using the exported file. rembg and UGT do not add that replacement background or edit the poster.
Use it again
Choose Run again to return to the photo form. Select the next image and start another run. The recording shows the form reopening; it does not show a second conversion.
If the result needs another try
An image with no clear subject can fail the transparency check. Try a photo where the subject stands apart from the background. A successful check confirms the file has the expected transparency; it cannot judge every fine edge for you.
This recipe uses CPU inference with the u2net model. See the current recipe for its inputs, output, limitations and original licences. Those facts stay with the recipe as it changes.
About this example
Recorded on an Apple silicon Mac with UGT already installed, its existing cache retained and its managed engine starting on demand. The full run took about two minutes on that machine. Your setup, photo and computer affect the time.
The photo is rembg’s upstream animal example. The MIT copyright and permission notice is retained. Read the on-screen captions.
