VideoToText

VideoToText is a powerful tool that allows you to download a video from YouTube or use a local video file, extract the audio, and transcribe the text using OpenAI's Whisper model. The project leverages Docker for easy setup and deployment.

Features

• Download video from YouTube or use a local video file
• Extract audio from the video
• Transcribe audio to text using OpenAI's Whisper model
• Save transcriptions in SRT and JSON formats

Technologies Used

• yt-dlp: A command-line program to download videos from YouTube and other video sites.
• ffmpeg: A complete, cross-platform solution to record, convert and stream audio and video.
• whisper: OpenAI's Whisper model for state-of-the-art speech recognition.

Project Structure

.
├── Dockerfile
├── docker-compose.yml
├── main.py
├── utils.py
├── README.md
├── /downloads  # Folder where the processed results will be saved
└── /files      # Folder where the videos to be processed should be placed

Getting Started

Clone the Repository

git clone https://github.com/vshloda/VideoToText.git
cd VideoToText

Build the Docker Image

docker compose build

Run the Docker Container

For processing YouTube videos:

docker compose run --rm app python main.py --url "https://www.youtube.com/watch?v=example"

For processing local video files:

docker compose run --rm app python main.py --file "files/video.mp4"

Changing the Model

To change the Whisper model used for transcription, modify the Dockerfile file. Update the line where the model is loaded:

# Define build-model arguments
ARG WHISPER_MODEL=small     # Change 'small' to 'medium', 'large', etc.

Available models include:

• tiny
• base
• small
• medium
• large

License

This project is licensed under the MIT License.

Acknowledgments

• FFmpeg
• yt-dlp
• OpenAI Whisper