---
license: other
tags:
  - comfyui
  - workflow
  - minimax-h3
  - refmod
---

# ComfyUI workflows

Public workflows from [COMFYUI 4 NOOBS on SKOOL](https://www.skool.com/ai-creators-8862/about). 

Each JSON in this repo is free to download.

## Workflows

| Workflow | File | Tutorial |
| --- | --- | --- |
| Create a MiniMax H3 RefMod | [CREATE-REFMOD-WF.json](CREATE-REFMOD-WF.json) | [YouTube Short](https://www.youtube.com/shorts/iL-8mupLCd0) |

## Create a MiniMax H3 RefMod

Builds a MiniMax H3 RefMod in ComfyUI from a reference image. The saved graph is set up as an identity RefMod (`Full Reference`, preset `identity_encode`) and writes the result as a bundle.

[![How to create a MiniMax H3 RefMod](https://img.youtube.com/vi/iL-8mupLCd0/hqdefault.jpg)](https://www.youtube.com/shorts/iL-8mupLCd0)

**Tutorial:** [How to use this workflow](https://www.youtube.com/shorts/iL-8mupLCd0)

**File:** [CREATE-REFMOD-WF.json](CREATE-REFMOD-WF.json)

**You need:**

- ComfyUI with MiniMax H3 support
- [ComfyUI-MiniMaxH3Mod](https://github.com/Luisacaotica/ComfyUI-MiniMaxH3Mod) (`MiniMaxH3RefModMasterExtract`)
- [rgthree-comfy](https://github.com/rgthree/rgthree-comfy) (group bypasser and labels)
- `minimax_h3_video_vae_fp16.safetensors`
- `minimax_h3_audio_vae_fp32.safetensors` when you turn the audio path on

**In this file:**

- **Image 1** is the active reference. It feeds `refs_image.ref_image_0`.
- **Image 2**, **Image 3**, **Image 4**, and the audio loader are in the graph and bypassed. Turn one on when you want another image, a video reference, or a voice.
- The video VAE loader is active. The audio VAE loader is bypassed with the audio input.

```