MiniMax open-sources H3-Image: 16MP images in 6 distilled steps, generation and editing in one model
RisingSayak · x · 2026-11-10
MiniMax has released MiniMax-H3-Image as open weights on Hugging Face — a unified image generation and editing model under a community license.
Key points:
- Direct 16-megapixel output at any aspect ratio from 4096×4096 (1:1) to 6272×2688 (21:9), no upscaler needed
- Distilled to 6 sampling steps (6 NFEs) with no classifier-free guidance, keeping compute practical
- Architecture: DiT with the H3 architecture minus audio layers, reusing the MiniMax-H3 text encoder unchanged
- One checkpoint handles text-to-image plus single/multi-reference image editing, with emphasis on precise instruction following
Available via API (global and CN platforms) and web apps at hailuoai.video / hailuoai.com.
Related event: MiniMax open-sources H3-Image: 16MP 4K images in about 10 seconds(20 posts)→
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
- MiniMax H3-Image 2-step LoRA released on Hugging Face with sample gallery — Illustrious_Row_9971 · 2026-11-11
- Magnific Studios Launches Animated Series 'Before the Show', Shares Behind-the-Scenes — CurieuxExplorer · 2026-11-11