An attempt to create a realistic ANIMA model.
Version 2.0
Version 2.0 was primarily tuned for the "er_sde" sampler, steps 12-16.
Run: Euler/simple (Beta), CFG 1, Steps 8-16
Text encoder: https://huggingface.co/circlestone-labs/Anima/tree/main/split_files/text_encoders
VAE: https://huggingface.co/circlestone-labs/Anima/tree/main/split_files/vae
Description
Base version 2.0, works optimally at sampler "euler" or "er_sde", CFG 2
FAQ
Comments (9)
The Model is great! Thanks! But there is no meta data in the example images T_T
It's really amazing. thanks you for this
can you uoload the base model for lora trainers
already
I’ve tested 2.0, and it’s fantastic. I tested it both (not the extra Turbo v2.0 version) with and without the Turbo LoRA, and both versions are absolutely amazing. The Base version with Turbo LoRa is better then the "Turbo V2.0" version. The realism and prompt adherence are incredibly good. It also handles fetish and NSFW content very well right out of the box. The only sliders I need are body sliders, and they work very well with it—though you do need to set the slider strength a bit higher. Thank you for this Checkpoint, i love it.
1.9.5 is much better then both v2 or v2 turbo in my testing.
Love this checkpoint - good prompt adherence and strong colors and contrast
while the level of realism is impressive, this model has a serious case of sameface syndrome
on base anima you can prompt for ethnicity and facial features but with this model one type of face is pretty much baked in
Yes, there is such a thing, but it's a sacrifice for the sake of greater realism.