manimator

!manimator

What is _manimator_?

manimator is a tool to transform research papers and mathematical concepts into stunning visual explanations, powered by AI and the manim engine

Building on the incredible work by 3Blue1Brown and the manim community, _manimator_ turns complex research papers and user prompts into clear, animated explainer videos.

🔗 Try it out:

• Gradio Demo:
• Or replace arxiv.org with manimator.hypercluster.tech in any arXiv PDF URL for instant visualizations!

Installation

[!IMPORTANT]

This project is built using the poetry tool to manage Python packages and dependencies. Download it from here to run this project or use the Docker image.

This project is dependent on the manim engine and hence has certain dependencies for running the engine properly which can be found here.

bash
git clone https://github.com/HyperCluster-Tech/manimator
cd manimator

Install Dependencies: poetry install

Activate the environment: poetry env activate

(If you're using a version before Poetry 2.0, you should use poetry shell)

Usage

After successfully installing all the project dependencies and manim dependencies, set the environment variables in a .env file according to the .env.example:

Run the FastAPI server:

poetry run app

and visit localhost:8000/docs to open SwaggerUI

Run the Gradio interface:

poetry run gradio-app

and open localhost:7860

Notes

To change the models being used, you can set the environment variables for the models according to LiteLLM syntax and set the corresponding API keys accordingly.

To prompt engineer to better suit your use case, you can modify the system prompts in utils/system_prompts.py and change the few shot examples in few_shot/few_shot_prompts.py.

🛳️ Docker

To use manimator with Docker, execute the following commands:

Build the Docker image locally. Then, run the Docker container as follows:

docker build -t manimator .

If you are running the FastAPI server

docker run -p 8000:8000 manimator

Else for the Gradio interface

docker run -p 7860:7860 manimator

{
  "status": "ok"
}

curl http://localhost:8000/health-check

{
  "scene_description": "Generated scene description based on PDF content"
}

curl -X POST -F "file=@/path/to/file.pdf" http://localhost:8000/generate-pdf-scene

{
  "scene_description": "Generated scene description based on arXiv paper"
}

curl http://localhost:8000/pdf/2312.12345

{
  "prompt": "Your scene description prompt"
}

{
  "scene_description": "Generated scene description based on prompt"
}

curl -X POST \
     -H "Content-Type: application/json" \
     -d '{"prompt": "Explain how neural networks work"}' \
     http://localhost:8000/generate-prompt-scene

{
  "prompt": "Your animation prompt"
}

curl -X POST \
     -H "Content-Type: application/json" \
     -d '{"prompt": "Create an animation explaining quantum computing"}' \
     --output animation.mp4 \
     http://localhost:8000/generate-animation

{
  "detail": "Error description"
}

Coming Soon

• Improved Generation Quality

• Video Transcription

• Adding Audio

• Chrome Extension

Limitations

• LLM Limitations

• Video Generation Limitations

License

manimator is licensed under the MIT License. See LICENSE for more information. The project uses the Manim engine under the hood, which is double-licensed under the MIT license, with copyright by 3blue1brown LLC and copyright by Manim Community Developers.

Acknowledgements

We acknowledge the Manim Community and 3Blue1Brown for developing and maintaining the Manim library, which serves as the foundation for this project. Project developers include: Samarth P, Vyoman Jain, Shiva Golugula, and M Sai Sathvik for their efforts in developing manimator.

Models and Providers being used:

• DeepSeek-V3
• Llama 3.3 70B via Groq
• Gemini 1.5 Flash / 2.0 Flash-experimental

Contact

For any inquiries, please contact us at [email protected] or refer to our website hypercluster.tech