Real Time Editing with GNRI Inversion 🍎⚡️

This is a demo for our paper GNRI: Lightning-fast Image Inversion and Editing for Text-to-Image Diffusion Models. Accepted to ICLR 2025. More details can be found in the project page.

The demo is based on SDXL. Improved results can be achieved with our FLUX version (examples in project page).

Input image

Enter your image description

Result

Enter your edit prompt

Examples

Disclaimer: The results may vary depending on the hyper-parameters. Errors may appear due to resource availability by HF. Performance may be inferior to the reported in the paper due to hardware limitation.