If the creator of NTRMix @tkio wants this taken down then I will.
Style based off of NTRMix 4.0. Obviously it is not 1 to 1 accuracy. Still trying to work out the settings of what is best for training style loras on anima.
Description
FAQ
Comments (5)
the new version for base has a tendency to mess up hands unfortunately
Shame, I didn't get those errors within my testing but I will try updating it in the future though.
@AkioAI I'm finding that one basic drawback of Anima is that it just doesn't do hands very well. It may not be your lora.
@Murf_ I see. It is surprising because usually updating a model means it is better than the previous model. I hope the author for the model updates it and fixes those issues if it is the model's fault and not my fault. 🤔
Try Euler A2 with Beta at 30-50 steps. Helps with hands a lot.