← sources

One .char model, Consistent face, body & cloths are now supported in Comfy(Custom node & workflows - Minimax H3 & Flux 2 Klein)

source: reddit (reddit:StableDiffusion)published 2026-10-03 13:13ingested 2026-10-04 02:37

retained text

I have been getting lot of request to publish .char support with ComfyUI, here is a custom node that I created: 

Node Repo: [https://github.com/omnichar/ComfyUI-Omnichar](https://github.com/omnichar/ComfyUI-Omnichar)

This repo includes few node, that let's you build a portable character model while keeping face, body & cloths. You can build a .char model once & use it across different workflows without a identity drift. 

**Features:** 

1. Build/encode character into .char
2. Decode .char built in Comfy or anywhere else
3. Decode into reference, character sheet & even LoRA adapter
4. SDK is public for future integration & improvements

**Nodes:** 

* **Decode Character** gives you the prompt, the reference images, a numbered contact sheet, and conditioning if you wire a CLIP
* **Character References Split** puts every reference on its own output(Dynamic)
* **Character Reference Latent** attaches every reference to the conditioning in one go
* **Character Reference** pulls a single reference by position
* **Encode Character** builds a `.char` from face, body and wardrobe references
* **Save Character** saves a .char into `models/characters`

**ComfyUI Workflows**

* [Build a character](https://github.com/omnichar/ComfyUI-Omnichar/blob/main/workflows/character_encode.json)
* [Generate images with Flux Klein 9B using a .char](https://github.com/omnichar/ComfyUI-Omnichar/blob/main/workflows/flux_klein_9b_image_char.json)
* [Generate videos with MiniMax H3 using a .char](https://github.com/omnichar/ComfyUI-Omnichar/blob/main/workflows/minimax_h3_char_video.json)

**Inputs**  
Checkout the guide in node repo for [references](https://github.com/omnichar/ComfyUI-Omnichar#guide)

**Prompt Guide**

* **Name your character**: Give your character a name so when passing prompt, I only have to say, **sia walking on the beach**.
* **Describe character features**: Encode all of the character features in description prompt & trigger your character with a name in generation prompt.
   * *Avoid describing same things in generational prompt.*
* **Handling Character drift**: *Each input refs should be unique, face should not have body or vice versa, same applies for clothing.*
* **Portability:** Once character is built, you can use the same character with only `Load Character` node. 

**Limitation:** Reference conflicts e.g. if two reference/input images has two different faces, it might conflict in generation, provide well cropped body & cloth images. Face images are crossed automatically by Sface.

**Related resources:**   
Omnichar repo: [https://github.com/omnichar/OmniChar](https://github.com/omnichar/OmniChar) (GPLv3)  
Community characters: [https://www.omnichar.org/characters](https://www.omnichar.org/characters)

Node submission is on the way. 

document fetched 2026-10-04 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1wwmywn/one_char_model_consistent_face_body_cloths_are