# Anime style enhancement

Animagine XL 3.1 is an upgraded version of the Animagine XL V3 series. It has been optimized on the basis of the previous work and can generate higher-quality anime images.

This is a VAE decoder fine-tuned based on stabilityai/sd-vae-ft-mse, specifically optimized for e621 image data. It employs a mixed loss function of MAE and MSE, calculating the loss in the Oklab color space to enhance image reconstruction quality.

A diffusion model focused on improving the quality of moe character generation, supporting high-resolution image generation

A diffusion model focused on improving the quality of moe character generation, supporting high-resolution image creation with optimized composition and anatomical structure representation.