This Lora was trained with Realism in mind. Enabling users to generate realistic furry characters with Anima.
This Lora and its versions have not been tested with any models other than Anima Base v1.
Version Information:
Version 1 is the sixth attempt at training this Lora, and the closest I've managed to what I was going for. The main issue from here is that the base Anima model itself has a very heavy digital art style bias because of the data it was trained on, which in itself didn't contain a whole lot of Furry. Meaning no matter how hard I train it, it likely won't be able to see its full potential until someone makes a proper furry fine-tune of Anima.
The SemiReal version is the fourth attempt at making an Anima furry Realism Lora, but unfortunately due to the training settings the style didn't end up as strong as I had hoped. The fur and scales for sure can be detailed, but overall it just looks SemiReal, but someone might find that useful so I thought I would post it anyways.
The Alpha version was my first Anima Lora I made, I clearly didn't use the right training settings, but it is able to create some half decent realism images so I thought I would post it up on here while I am continue training.
Description
This is the fourth attempt I have made in trying to train this. Some times it looks not half bad for realism, but most of the time it ends up looking more semi-real than anything. I think it has to do with the fact that I tried training this one using Prodigy instead of AdamW8Bit. But since someone might find it useful I might as well throw it in here. The tag I am training on "photorealistic" might be diluted by information trained into the model itself, so if I can't get this in the next couple attempts I will probably try a unique tag instead.