MiniMax H3 video reference pack

A RefMod replaces LoRA training for MiniMax H3. Instead of training a LoRA, you save a reusable reference of a character, a video or a sound in a few minutes, then load it whenever you generate.

The RefMod pack has four workflows: three that create RefMods (from images, video or audio) and the main MiniMax H3 RefMod workflow that uses them. This workflow turns a short video clip into a video RefMod that captures its appearance and motion.

Great for:

• Reusing a dance or movement across AI character videos
• Keeping a scene's look consistent
• Building a library of motion references

How it works

1. Put your clip in the input folder.
2. Run the workflow to save the RefMod.
3. Load it in the MiniMax H3 RefMod workflow.