UNET only trained model on small dataset. Due to presence of old images in dataset resembles averaged over time nowa's style.
To my eye on anima works worse than on illustrious
If you are experiencing resemblance issues on SD-Forge-Neo try changing in settings option Stable Diffusion -> Compability -> Try to reproduce the results from external software -> Comfyui
Description
removed some lowres images
noise offset
step ladder function used during training