This is a style model that does art like the MARVELOUS! game company conception plus, this was trained on conversations with every girl in the game. The game only has convos (to my knowledge) with only 1 girl at a time, upper body only so remember that when generating. Tends to work best at 0.6. This model was trained on NAI anime-full-pruned and generated in CitrineDreamMix. Made with the Kohya tool for AI Training through google colab. Any suggestions to improve or recommended things for the data set would be appreciated.
Description
Trained on 143 screenshots during my time playing the game, don't know why but has difficulty generating pics that look like the gemini maiden. Specializes in long images with a single character, probably a good tool for visual novel test. For a future update rather than adding more images simply adding more image details about the backgrounds (cause at the moment they all look similar), and some of the more unique tags for features and accessories will help.