ChatWithYT is a tool that allows users to chat with YouTube videos. Users provide a list of YouTube URLs, and the system processes the videos by downloading their audio, converting it into transcripts using Whisper, storing the data in a vector database, and enabling a conversational interface to interact with the extracted content.
Ensure the following dependencies are installed before proceeding:
- Install Ollama
- Download the required Ollama models:
ollama pull gemma3:4b ollama pull nomic-embed-text
- Install FFmpeg (for processing audio):
- Windows: Install FFmpeg
- macOS:
brew install ffmpeg - Linux:
sudo apt install ffmpeg(or use your package manager)
# Clone the repository
git clone https://github.com/Ashenoy64/ChatWithYt.git
cd ChatWithYt
# Set up virtual environment
python -m venv .venv
source .venv/bin/activate # On Windows use .venv\Scripts\activate
# Install dependencies
pip install -r requirements.txtCheck Settings.py to configure the system:
WHISPER_MODEL: Defines the Whisper model used for audio transcription. Options include 'tiny', 'base', 'small', 'medium', and 'large'.VECTOR_DB_NAME: Specifies the name of the vector database where transcript embeddings are stored.DOCUMENT_SPLIT_OPTIONS: Configures how text is split into smaller chunks, including chunk size, overlap, and separator priorities.SHARE_CHAT_INTERFACE: If set toTrue, allows the chat interface to be accessible over the internet.EMBEDDING_MODEL,LLM_MODEL,CHAT_MEMORY: Define the embedding model, language model, and chat memory storage for conversations.
Run the script with one or more YouTube URLs:
python main.py https://www.youtube.com/watch?v=<VideoId>For example:
python main.py https://www.youtube.com/watch?v=dQw4w9WgXcQURLS(positional argument): One or more YouTube video URLs that you want to process.-h, --help(optional argument): Displays help information about usage and available options.
Feel free to fork the repository and submit pull requests!
This project is licensed under the MIT License.