/vt/ - Virtual Youtubers

Shiki Miyoshino (summer) LoRA
There is a style bias because it's mostly model shots. On my NAI test model (block merge of based 66 67 and 64) a beach consistently makes it into the background until I bring the weight down to around 0.8. On EF, the ears can be annoyingly inconsistent, weight 1.2 helps, but it doesn't help the jacket oddly enough. I might try a fluffyrock version. Testing this shit on forge is annoying because it likes caching loras by name until exit.
RAPE RAPE RAPE SEX RAPE SEX RAPE https://mega.nz/folder/dfBQDDBJ#3RLMrU3gZmO6uj167o-YZg/file/lXwUkLwC
Also here is the dataset if someone wants to make an SDXL LoRA https://mega.nz/folder/dfBQDDBJ#3RLMrU3gZmO6uj167o-YZg/file/dD5iGBbS