Tests were conducted to ensure:
1) That the real-life images would not lean towards the anime style;
2) That in cases where the emotions are similar, the generated images would have distinct facial differences for the characters.
Description
FAQ
Comments (3)
Which filterbypass LoRA and at what strength did you use to get these expressions working? Thanks
10 Strength
0 info about the tags you trained, about the Lora's strength, only 1 sampler where vaguely something happens, no XY comparison, etc. Impossible to know what's that worth.
I mean thanks for taking the time to train something. But don't upload it if you don't take the extra time to explain and show clearly how this influences the base model.