# Minimax H3 Refmod Workflow — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/a93e602ba45f2455

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-10-12 19:07 ingested 2026-10-13 02:40

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1x4auud/minimax_h3_refmod_workflow/)

## retained text



```
https://preview.redd.it/f7ytj63103vh1.jpg?width=2996&format=pjpg&auto=webp&s=4f5ff97fae270d5b912fcb4193dd75b865604dce

Since my last post has been deleted unfortunately (due to a mega-link), here we go again.

I've spent quiet some time now with Refmods and this workflow, and I'am still baffled how easy it is to create a character identity with very low effort. This is a no-brainer if you don't want to or can't train a Lora, all you need is good pictures and a good voice sample, the refmod-training is done in literally seconds (depending on your GPU of course).

I did my very best to make this workflow easily understandable for even beginners, nevertheless here's some explanation:

The workflow actually consists of 2 workflows, one for the refmod creation (the nodes in the top right of the picture, that's a different workflow but i just wanted one picture) and the main workflow for creating a video with the created refmod file which is all the rest.

If you just wanna use a refmod file and no extra images, videos or audio, select the image/video/audio nodes and press CTRL+B to deactivate them. Please watch the video from Dainamo for an detailed explanation, I've made some notes but for further understanding just watch it. I just wanna give you an easy workflow, that is of course fitted to MY needs, do whatever you want to it.

Here's some color-explanations:

\- GREEN is all sources for the final video so to say: images/video/audio/the prompt (deactivate them if you just wanna use a refmod-file, but you can also reference clothings or whatever this way)

\- TURQOUISE is all relevant video-settings

\- TAN is the refmod-nodes (except the 4 notes)

\- RED is all the models

\- PURPLE is all the loras (turbo-lora is in light-blue and another lora above in grey for whatever)

\- BROWN is comfy-kitchen and the spectrum-node (not activated, transparent-purple)

\- GREY is irrelevant

Here is the link to the original tutorial from Dainamo (thanks again!!): [https://www.youtube.com/watch?v=2K6-OtV\_Vbc](https://www.youtube.com/watch?v=2K6-OtV_Vbc)

And here are the workflows: [https://getshared.com/d/vrI9CYLk](https://getshared.com/d/vrI9CYLk)

I hope you enjoy the workflow! :)
```



document fetched 2026-10-13 00:33 · canonical https://reddit.com/r/StableDiffusion/comments/1x4auud/minimax_h3_refmod_workflow
