MiniMax H3-Image 2-step LoRA released on Hugging Face with sample gallery
Illustrious_Row_9971 · reddit · 2026-11-11
A Reddit user shared a 2-step distilled LoRA for MiniMax's H3-Image model, hosted on Hugging Face (akhaliq/minimax-h3-image-pdmd-2step), cutting sampling down to just two steps. The attached gallery shows results across ukiyo-e illustration, photoreal animals, anime portraits, sticker sheets, and editorial-style photography, with stable quality and good detail retention.
More from Multimodal
- SenseNova 6.8 Code2Video generates videos via editable code, fully open-sourced — aftahi_ai · 2026-11-11
- MiniMax H3-Image vs. Pruna AI compressed version, same prompt and seed — CutLongjumping8 · 2026-11-11
- Seedance 2.5 generates a luxury perfume commercial for a fictional fragrance — aftahi_ai · 2026-11-11
- BaoCut Explains Virtual Streamers Like Shadow Puppets — 3 Steps to Add One to Your Videos — dotey · 2026-11-11
- Magnific Studios Launches Animated Series 'Before the Show', Shares Behind-the-Scenes — CurieuxExplorer · 2026-11-11
- 'Cara Love the Adventurer': AI-Generated Short Film Trailer Released — noLOOPai · 2026-11-11