Improvements to SDXL in NovelAI Diffusion V3
arXiv:2409.15997
Abstract
In this technical report, we document the changes we made to SDXL in the process of training NovelAI Diffusion V3, our state of the art anime image generation model.
14 pages, 8 figures