Experimental local background remover
This experimental page is excluded from public launch because first-run model downloads and device compatibility are not consistently reliable.
How local background removal works
The page loads the Apache-2.0-licensed ISNet General Use ONNX segmentation model through Transformers.js. The model predicts an alpha mask, which is applied to the original pixels before a transparent PNG is created.
What to expect
The first run downloads a large machine-learning model and can take time on a slow connection. Processing speed depends on the device. Clear subjects and visible boundaries generally work better than transparent objects, hair-like detail, heavy motion blur, or multiple overlapping subjects.
Privacy and licensing
The selected image is processed in the browser and is not intentionally uploaded. Model files are downloaded from Hugging Face. ISNet General Use ONNX is Apache-2.0 licensed; Transformers.js is provided by Hugging Face under its applicable open-source license.