← sources

The new Minimax H3- Image Model seems rushed and buggy.

source: reddit (reddit:StableDiffusion)published 2026-11-15 16:45ingested 2026-11-16 02:37

retained text

I will be honest, i have tried the Minimax H3 Image when it dropped on comfy day 0 and the images were amazing, but lacking detail despite being very high resolution.  
I thought maybe upping the MP would fix it but it didn't and then there came the anatomy issues as well.  
I checked some posts about this, got the p-h3 vae and all but it still didn't fix it.  
So checked another post talking about how the edit mode is better and i was happy but that was actually FAR WORSE. It lacks character consistency, its too prompt sensitive in a bad way, and its just not good.  
Now i do not know if Hailuo is at fault or Comfy just released some poorly fine tuned version, is anybody else feeling the same way or.  
Cause even the templates they released are half nerfed and I don't know it just feels bad, like very very bad.  
If it at least nailed the editing capabilities like its video model could do I honestly would not really care about the other stuff, but it seems finding the right edit model in this open source space is almost impossible.

I hope krea 3 edit or Ideogram 4.5 would be good ( If they ever release them that is )

document fetched 2026-11-16 00:29 · canonical https://reddit.com/r/StableDiffusion/comments/1xx06c8/the_new_minimax_h3_image_model_seems_rushed_and