Repolode
Submit
omnichar

omnichar/OmniChar

Open-source studio and portable .char file format for keeping AI characters consistent across image and video models, locally or via hosted APIs.

View on GitHubWebsite
Stars
556
Forks
74
Open issues
0
Last commit
5 days ago

What is it?

Omnichar builds AI characters that keep the same face, body, clothing and voice across shots and models. A character is packaged as a single portable .char file that holds reference photos, an identity fingerprint, a locked description and optionally a voice clip. Generation can run on your own NVIDIA GPU, inside ComfyUI through a companion custom node, or through hosted models via fal.

The studio is a Python app served from one local port, with a node canvas for prompts, model nodes, LoRA training and video editing. Supported models include FLUX.1, FLUX.2, Krea 2, Z-Image Turbo, MiniMax H3 and LTX-2.5. Some models use the reference images directly, while others train a small LoRA adapter from the same photos. Each take is scored for continuity against the character.

LoRA training runs locally on stills or short clips, and the trained .safetensors files are loaded straight into the generation nodes. The project is GPL-3.0 licensed, and model weights carry their own separate licences.

Key features

  • 01Portable .char files bundle reference photos, identity fingerprint and locked description.
  • 02Local generation and LoRA training on NVIDIA GPUs, with a CPU fallback.
  • 03Supports FLUX.1, FLUX.2, Krea 2, Z-Image Turbo, MiniMax H3 and LTX-2.5.
  • 04Hosted models via fal, with the same character applied to API requests.
  • 05Works as a ComfyUI custom node that reads the same .char format.
  • 06Each take receives a continuity score against the character.

Use cases

  • Keeping one character consistent across a series of images or a short film.
  • Training a LoRA on your own photos or video clips without a cloud step.
  • Generating character video with synchronized sound using MiniMax H3 or LTX-2.5.
  • Sharing a character file between ComfyUI workflows and a standalone studio.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/omnichar/OmniChar
  2. 2

    Change into the core directory

    cd OmniChar/core
  3. 3

    Install everything (macOS/Linux)

    ./webui.sh --install --extra all
  4. 4

    Start the web UI and open http://127.0.0.1:8848

    ./webui.sh

Things to know

  • Needs Python 3.11+ and, for local generation, a capable GPU; Krea 2 inference needs about a 40GB card.
  • Some hardware paths such as Apple Silicon and AMD ROCm are untested.
  • Hosted model options depend on external APIs rather than running fully locally.

Topics

#ai-art#generative-ai#lora#diffusion-models#character-generator#video-generation#comfyui#ai-film#ai-filmmaking#ai-video

FAQ

What is OmniChar?
Open-source studio and portable .char file format for keeping AI characters consistent across image and video models, locally or via hosted APIs. Omnichar builds AI characters that keep the same face, body, clothing and voice across shots and models. A character is packaged as a single portable .char file that holds reference photos, an identity fingerprint, a locked description and optionally a voice clip. Generation can run on your own NVIDIA GPU, inside ComfyUI through a companion custom node, or through hosted models via fal.
Is OmniChar free to use? What license does it have?
OmniChar is open source under the GPL-3.0 license. Check the license terms before using it in commercial products.
Is OmniChar still maintained?
Actively maintained. The last commit to OmniChar was on Oct 16, 2026.
What can I use OmniChar for?
Keeping one character consistent across a series of images or a short film. Training a LoRA on your own photos or video clips without a cloud step. Generating character video with synchronized sound using MiniMax H3 or LTX-2.5. Sharing a character file between ComfyUI workflows and a standalone studio.
How do I install OmniChar?
Clone the repository: git clone https://github.com/omnichar/OmniChar Change into the core directory: cd OmniChar/core Install everything (macOS/Linux): ./webui.sh --install --extra all Start the web UI and open http://127.0.0.1:8848: ./webui.sh
Where can I get OmniChar?
The source code and installation instructions are on GitHub: https://github.com/omnichar/OmniChar

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects