AuTeVi - Automatic Text to Video for multimedia distribution

A powerful solution, driven by LLMs and a Multi-Agent System, designed to automate the creation of videos, presentations, and documents for various use cases.

Advantages

• Automated Content Generation: Create exceptional contents for content makers just by some clicks.

• Quality Assurance: Ensure the quality of the contents and products with our well-designed flow containing best AI agents.

• Cost-Effective Productions: Reduce operational costs by automating production tasks, eliminating the need for manual content creation, and streamlining processes.

• Analysis and Feedback : Leverage data-driven insights to refine and improve content through continuous analysis and feedback loops.

• Easy editing: Customize and modify outcomes seamlessly, allowing you to tailor content to your specific needs with minimal effort.

Installation

It's easy to install this project by the following commands (ensure that you have docker):

git clone https://github.com/hoangbros03/AuTeVi.git
cd AuTeVi
# Now we have 3 folder: frontend, backend, ai

# Create mongo as DB
# There are other ways, but using Docker is the simplest way
docker run --name mongodb -d -p 27017:27017 mongo

# Run backend with java spring boot
./mvnw spring-boot:run

# Run AI component
cd ../ai
pip install -r requirements.txt
pip install -e .
fastapi dev src/main.py --port 6969

# Run frontend run with next.js
cd ../frontend
npm install
npm run dev

Then we can start by go to localhost:3000

Environment Variables

Please note that due to an extremely limited time, we now store the api keys on a file inside AI component. Please change if needed. The API keys are:

JSON2VIDEO_API = 'XXXXX' # To convert JSON to video
TAVILY_API = 'XXXXX' # To search images & information
GEMINI_API = 'XXXXX' # To call the LLM
IMGBB_API='XXXXX' # To upload image online
DB_LINK = 'XXXXX' # Link to MongoDB
EMBED_MODEL_ID = 'XXXXX' # To use RAG powered with AWS Bedrock

Demo

Acknowledgements

This quick prototype was developed during the 24-hour coding session of the Swin Hackathon 2024. We aimed to create a functional, feasible, and visually appealing prototype. A big thanks to our teammates and the organizers!