/vt/ - Virtual Youtubers

which settings for lora extraction are better?
dim 32 is too low, Lui looks wrong
dim 320 feels too big, file is 480MB
dim128-conv_dim64 is 230MB
dim256 without conv_dim is 290MB