Skip to main content
By the end of this page you have an image of your own character, rendered in ComfyUI. You need a working ComfyUI with Python 3.10 or newer, and the FLUX.2 klein 9B model, text encoder and VAE files.
1

Install the node

Restart ComfyUI.
2

Build a character

Download character_encode.json and drag it onto the ComfyUI canvas. Load your own face, body and wardrobe photos into the Load Image nodes, write a short description in Encode Character, and set a filename in Save Character. Queue it.The .char is saved to ComfyUI/models/characters/.
3

Open the FLUX.2 Klein workflow

Download flux_klein_9b_image_char.json and drag it onto the ComfyUI canvas. Point the model, text encoder and VAE loaders at your FLUX.2 klein 9B files.
4

Pick your character and run

In Load Character, choose the .char you saved in step 2. Write a prompt for the scene and queue it. Decode Character resolves the references, and two Character Reference Latent nodes feed them to the positive and negative conditioning.
Already have a .char from Omnichar Cloud or Omnichar Studio? Put it in ComfyUI/models/characters/ and skip the build step.

​
Next steps

Workflows

MiniMax H3 video, and giving a character a voice.

Nodes

Every node, with its inputs and outputs.