This version is kind of unstable and we haven't tested on all of it's generation results. At least for now, we feel like the style are overfitted and we should probably train it locally for better results from now on. Also we might switch to train on the Animagine v4.0 model instead of illustrious.
Next up we are fixing:
Full body generating issues
Eye color and eye fixure
Style control for individual characters(such as not able to generate a specific character as given)
The issue with Anime Resolution 4x not working with the upscaler under this model
Enabled for higher resolution generation(up to potential 1536x1536)