Skip to content
かみもと · 本地AI视频工作流·· 2026-10-20AI score58

Author trains a style LoRA "SepiaBlue Luminous" for MOSS-Image and releases it publicly on Civitai

【MOSS-Image】画風LoRA「SepiaBlue Luminous」を作ってCivitaiで公開しました!

AI briefing

The author trained a style LoRA "SepiaBlue Luminous" for the local image generation model MOSS-Image and released it publicly on Civitai, with the trigger word qsv2style. Training used 100 ChatGPT-generated images sharing only the style, with different characters and scenes, based on MOSS-Image-Base, at 1024 resolution, rank 16, learning rate 1e-4, for 1000 steps, and the 400-step version was ultimately chosen.

Open original note.com