Redditor trains Seedream 5.5 style LoRA for MiniMax H3 in 6K steps, works on video too

NebSH83 · reddit · 2026-11-20

Reddit user NebSH83 released an open-source style LoRA that transfers Seedream 5.5's photographic look — lighting, contrast, and film grain — to MiniMax H3 Image.

Available on Hugging Face and Civitai, with the author soliciting feedback across subjects and lighting conditions.

Related event: Seedream-Style LoRA for MiniMax H3 Image Tops HF Trending(3 posts)→

Original post →

More from Multimodal

Multimodal channel →