First and foremost, and most importantly: ALL SAMPLE IMAGES WERE GENERATED USING TEXT-TO-IMAGE ONLY WITH THE ANIMA BASE v1.0 CHECKPOINT. No Hi-Res Fix, no ADetailer, no upscaling whatsoever.

That being said, this LoRA is part of an attempt to create a realistic model for the Anima family of models. I was going to label it as a beta, but it's not even quite there yet. However, with the right combination of realistic tags—and perhaps paired with a semi-realistic or realistic checkpoint—things could get interesting.

I haven't tested it with any checkpoint other than the official one and another one I'm currently cooking up. I have no idea if it works with other models, but the closer it is to Anima Base v1.0, the better it will perform.

LoRA weight? 0.9 / 1.0. Personally, I almost never go below 1.0.

Which tags do I use? realistic, photorealistic, hyperrealistic, amateur photo, 8k, etc. Give it a try...

I hope you like it and have as much fun with it as I did. Looking forward to seeing your crazy creations on the wall!

HaDeS