# Model recommendation for a motion transfer — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/ba1e9819591aa491

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-10-10 15:30 ingested 2026-10-11 02:39

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1x2i4ax/model_recommendation_for_a_motion_transfer/)

## retained text



```
Sup guys, ive been looking into motion transfer for a while but i cant manage to find a good model (or workflow) for motion transfer...

I already tried:

H3 - it was taking more than 30min to generate a 720p24 video using 4 steps on pro6000 and 96gb of ram, other than that it was the best one

Scail2 - couldnt get the realism right, there was just always some grain either on hair or in the background, also i had problems with eyes

Wan22 - masking issues (could fix that) and same as scail i just couldnt get the realism right

Before that i tried some non-local alternatives:

Kling, seedance - identity problems since i couldnt use a lora (tried some harder characters to generate)

Thoughu i havent checked out ltx2.3 and 2.5 yet 

So im trying to seek help between more experienced users here :D since i dont know much yet, im just starting up and i only know some basics. Id really appreciate some advice, tips and tricks so i know how to finally make a good workflow, or even whole wfs so i can learn from them... 

Thanks for any advice
```



document fetched 2026-10-11 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1x2i4ax/model_recommendation_for_a_motion_transfer
