Ayase Yuu of Phase Connect JP
YT: https://www.youtube.com/@ayaseyuu
Flux Dev & Schnell - Same dataset as SDXL. Weights 0.75 - 1 recommended.
SDXL V1.0 - Finally created an SDXL variant with post-debut images. Trained using FlatpieceCoreXL + FlatpieceXL VAE. 0.6 weight recommended.
Tattoo has a better chance of appearing correctly using "abdomen tattoo"
V2.0(SD1)- Trained on higher quality images from pre-debut trailer. Tattoo and monocle are bugged and don't properly generate.
V1.0(SD1) - Trained on pre-debut twitter images. Tattoo isn't exact. Has a tendency to include another person on the image.