A week doing training and merging to obtain a first good base to continue improving the model.
Description
FAQ
Comments (5)
Holy shit this is good, The prompt adherence is next level
This is relatively solid.
Still seems that there's still no Anima model that understands spatial relevance.
Very hard to make characters not giant in comparison to their location, and it's bleeding me dry
Surely some kind of LoRAs to circumvent the problem will come soon.
Turbo LoRAs seem to not cut it.
very good data, probably with a good vl over diferente positions images can be enough to fix it. I will study it.
Could you show me one example to check it in new versions?
@frakerkill_ Your main image of the girl wearing the blue kimono is actually a great example
It looks as if she were to stand up within the room, then she'd be gigantic
I'd be surprised if you could actually fix it
Just found foreshortening slider LoRA for regular Anima
Going to try it soon, at like -1.00 or -.0.50 strength to see if it works