Remove image background, use it like this
Select a PNG, JPEG, or WebP image and click Remove Background. Compare the original and processed results, then save as a transparent PNG download. AI models are downloaded from this site and photos are processed only on the device.
Example usage
Try using photos of people or products that stand out clearly from the background. The area that appears as a checkerboard pattern is a transparent area and the downloaded image does not include the pattern.
Please note
Supports up to 15 MB and 16 million pixels. The results are saved within 2,048px on the long side. On first run, fast mode loads about 16MB and precise mode loads about 188MB. Fine mode is recommended for PC and allows you to adjust the residual background and edges after processing. Separation quality may be poor for hair, fur, translucent objects, or complex backgrounds.
Step-by-step instructions
- Select an image and choose Fine or Fast mode.
- Press Remove Background to load the model and process the photo.
- Clean up the residual background and adjust the values inward to the edges to see the results.
- Save the file with the current correction value reflected by transparent PNG download.
Select fine mode and fast mode
Precise mode uses a large U²-Net model and downloads approximately 188 MB, including the execution engine, when first run. Fast mode uses U2NetP and requires approximately 16 MB initially. The model only requests photos to be processed and the photos are not sent to the server. Both modes may leave some errors if the boundaries are ambiguous, such as complex backgrounds, transparent glass, or fine hair.
| Settings | Recommended Situation |
|---|---|
| Precision Mode | Processing complex subjects on PC/Wi-Fi |
| Quick Mode | Mobile/small download capacity |
| Increase residual background cleanup | Remove faint background, be careful of loss of fine hair |
| 1~3px inward | Reduce smudged borders, be careful of loss of thin structures |
Frequently Asked Questions
Is the checkerboard pattern also saved?
No. This is a screen display to distinguish transparent areas. Transparency is stored in downloaded PNGs.
Are photos used for servers or AI training?
Photos are processed only within the browser. This site does not upload input photos or store them as training data. Access records for receiving model files can be handled by your hosting provider.
What should I do if my hair disappears?
Clean up the residual background and lower the value inward or press View without correction. Because it is recalculated from the original mask, corrections do not accumulate over repeated adjustments.
Can I receive very large photos in their original size?
Currently, input supports up to 15MB/16 million pixels, and output supports up to 2,048px on the long side. The original file is not modified.
Written and operated by: Gu Ja-hyeok · Content review: 2026-09-30 · Error report