The Large Body Language Model is an interactive installation made in the context of a seminar at the Einstein Center Digital Future in Berlin.

Run the Installation

0. Requirements

• Python 3.10 or higher
• A dedicated GPU (Cuda is recommended, but Apple Silicon is also supported)
• Webcam
• A Display or Projector
• roughly 10 GB of free disk space / min 16 GB RAM

1. Download the repository

You can download the repository manually or by using git:

git clone https://github.com/LangeRobert/lblm

2. Install the requirements

All required Python packages can be installed using pip. Make sure you are in the root directory of the cloned repository and run:

pip install -r requirements.txt

3. Start

To start the installation, run the following command in the root directory of the cloned repository:

python -m lblm

The first execution will take some time, as the model will be downloaded and cached. After that, the installation will start and you will see a window with the webcam feed and the displayed animation.

4. Interact

You can interact with the installation by moving your body in front of the webcam. Try things out. Have fun. Think about the first interactions with ChatGPT and how it felt to talk to a machine.

Thanks

Huge thanks for the support and the great seminar to Michelle, Florian and Friedrich. Also thank you to the wonderful people that have built the great software used in this project, primarily: MediaPipe, The HuggingFace Community, Panda3D and Mixamo.

Legal

Please note that this project is for educational/artistic purposes only and not for commercial use. Look at the LICENSE file for more information.