# Minimax H3 Help with Pose transfer without merging/bleeding in characters. — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/dbb9947925801293

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-09-06 16:15 ingested 2026-09-07 07:18

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1w90ekp/minimax_h3_help_with_pose_transfer_without/)

## retained text



```
i've been trying for ages, did everything, asked every LLM, i might be dumb but i've seen people doing it, my question is simple:

how do i make <Subject 1> from <Picture 1> do the pose in <Picture 3> in in ref2va (reference to video) ? 

everytime i try, <Subject 1> shifts into looking like <Picture 3> while doing the pose.

please write me the prompt example and a video would be appreciated.
```



document fetched 2026-09-07 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1w90ekp/minimax_h3_help_with_pose_transfer_without
