Lora\'s Tuning

I didn\'t like the output of the previous Lora creation (below), which had a half-realistic face, and also a rather dark face (in terms of construction and coloring...), so I tried tuning it. Ta. https://www.aipictors.com/works/466884/ In the process of Lora\'s operation, there are 12 steps (called blocks) of processing in SDXL, and I tried to leave only the minimum processing that brings out the characteristics. - sd-webui-lora-block-weight: Now find only the parts where long hair, color scheme, and fox appearance are processed. - Written with sd-webui-supermerger: (From April to the present (2024/9/28), there is a bug that prevents you from writing BlockWeight to single Lora, so revert to commit hash 97030c89bc94ada7f56e5adcd784e90887a8db6f and apply). Now you don\'t have to apply weights every time (sweat) generated by pony diffusion V6(trained by PurpleSmartAI) If applied, it may be possible to tune Lora, which is problematic because even the picture changes when applied.