Remove photo background
Remove the background of the person portrait photo from the browser and save as transparent PNG. The first run is for public examples; your photo is processed on the device.
Person portrait photo
Sample photo is being prepared.
Result with transparent background
Can immediately remove background with Public example.
Click ‘Background removal’ to view the result.
Boundary adjustment
Increasing the value makes the faint pixels at the boundary more transparent. It is not an editing feature that recovers parts incorrectly separated by the model.
Share result
Share the tool link and result text together. No URL is included in the result. Check the below content and then share it.
How to remove the photo background
When opened, 512×512 public example person photos are displayed. Click ‘Background removal’ to create a transparent PNG of the same size. Select your photo to use the example instead, and the photo content and filename are not included in the URL or analysis.
This page predicts transparency using MODNet person mapping model and generates a transparent PNG in the browser’s canvas. The longest result length is 2,400px. No original file is overwritten, and no metadata about the photo’s location or shooting is added to the new PNG.
The model is designed to estimate smooth boundaries like human hair and clothing. The boundary may be inaccurate in cases with single-color backgrounds, small subjects, products, animals, multiple people, or overlapping objects. Check the opacity preview and save with adjusted boundaries if needed.
Frequently Asked Questions
Will it completely remove product photos or complex scenes?
This tool uses the MODNet model tailored for character separation. Hair and semi-transparent borders may vary depending on the photo, and results may not appear in scenes with overlapping products, animals, or multiple people. Check the edges on the checkmark background before saving.
Is the photo sent to the server?
No. Images are processed by the model in the browser. On first execution, approximately 6.6MB of model and about 26.9MB of WASM executable are downloaded from this site. The gzip response on GitHub Pages as of 2026-10-04 is approximately 12.0MB for both files combined, with smaller file and transmission volumes being separate. Actual transmission volume may vary depending on browser cache and response compression. Image files are not included in network requests. Temporary results and images are deleted when processing is interrupted, initialized, or page moved.
Is the original size and photo information preserved?
The original file is not changed. To reduce memory usage, the longest side of the result is up to 2,400px, and the original metadata such as location and shooting info is not included in the PNG file. The result stores the background as transparent.
Calculation criteria and reference materials
- Xenova/MODNet — model for removing person photos backgrounds
Person portrait photo — Matt and Apache-2.0 model cards. Does not guarantee accuracy for separating all objects and complex scenes.
- Hugging Face Transformers.js — remove browser background
API for the image-to-alpha background removal pipeline executed in the browser.
- MDN — OffscreenCanvas.convertToBlob()
Encode canvas data inside the walker as a PNG file
Verification of calculation criteria: · Calculation·verification principles · Report errors