← sources

MiniMax H3-Image one-checkpoint Edit

source: reddit (reddit:StableDiffusion)published 2026-11-13 09:42ingested 2026-11-14 02:38

retained text

 MiniMax H3-Image swapping the subject, coat, and pose in a 6-step edit (with scale and grade locked)

  
**Specs**: RTX 3090, int8, \~60 s  
**Workflow**:   
UNETLoader(int8) → KSampler(euler/simple, 6 steps, cfg 1)  
CLIPLoader(nvfp4, type minimax)  
VAELoader(image\_vae\_fp16) → VAEDecode  
LoadImage ×2 → MiniMaxH3ImageConditioning(prompt, refs=<Picture 1/2>) → ConditioningZeroOut(neg)





document fetched 2026-11-14 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xv3ogw/minimax_h3image_onecheckpoint_edit