Trained on 90 screenshots of them from the anime. It is now better because of a new dataset, and It was trained on LoCon instead Of LoraCredit goes to Cybermeow for the program used to gather the dataset<a target="_blank" rel="ugc" href="https://github.com/cyber-meow/anime_screenshot_pipeline">https://github.com/cyber-meow/anime_screenshot_pipeline</a>

Echidna (Re:Zero)

better model with new more diverse dataset and trained using LoCon

Files with SHA256: 42b9e647d955823d6a4c394cf3549ac1cf6586f160276d198e3687a55ce6ad03