omnichar /OmniChar

About OmniChar

Omnichar Studio is a free, open-source app for AI image and video generation. Build a character once, and the same face carries across shots and models. It saves that identity in a portable .char file, or trains a LoRA (a small add-on model) for you. Some hardware paths are untested, and hosted models need paid API keys.

Best for: Solo filmmakers and AI artists who need the same character to look right in every shot.

• Saves a character's face, body and wardrobe into one portable .char file
• Runs generation, LoRA training and video on a node canvas
• Scores each take out of 100 and flags the worst frame in a clip
• Trains LoRAs on your own GPU; finished files land in models/loras/
• Runs local models on your GPU or hosted models through fal API nodes
• Downloads the diffusion model, VAE and text encoder with progress shown

1. 1
   Install Python 3.11+, then clone the repo: the command below

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

   Copy
2. 2
   Go into the core folder: cd OmniChar/core
3. 3
   Install the engine, runtime, trainer and UI: the command below (Windows: the next command)

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

   Copy

   .\webui.bat --install --extra all

   Copy
4. 4
   Start the web UI with ./webui.sh and open http://127.0.0.1:8848
5. 5
   Or install from PyPI with the command below, then run inline-studio

   pip install -r requirements.txt

   Copy