MOEMOSS-Base

MOEMOSS-Base is my anime-style fine-tune, made for creating colorful anime illustrations. I wanted a model that brings vivid color to character art, portraits, and illustrated scenes.

It was trained on 10,800 high-quality, high-scoring anime-style images, including 800 images I mixed into the dataset to steer the model toward my personal preferences. I find that the mixture created a fine balance of style that plays nice with different kinds of Style LoRa.

This is the foundation for future updates. I plan to expand the training dataset over time, building on this release while keeping the colorful anime look that I made it for.

Sample settings: 50 steps, CFG 3, Euler sampler, sgm_uniform scheduler. The checkpoint showcase images use MOEMOSS-Base alone, without any LoRAs. Plays nice with the Turbo LoRa's

~~Crafted with Love~~