This is a bit outdatad now, please check out the LoRA I made.

Making models can be expensive. Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee ☕

Textual Inversion trained on One Piece anime screenshot (Wano arc and Stampede movie). Seems to be working fine on most anime models, mainly AnythingV3 and Midnight Mixer (all of them). It also works with the One Piece character embeddings I made.

Versions 5000 and 22000 seem to be working better with male characters, but it might just be RNG. 5000 also seems to be less "messy" at times and gives a softer effect (less anime). The Yamato pic in the gallery was made with 5000.