openai_whisper_transcribe_yt_videos

YouTube Video Transcriber

This project is a Streamlit-based application that allows users to download audio from YouTube videos, transcribe them using OpenAI's Whisper model, and display the transcriptions with pagination.

Features

• Download audio from YouTube videos
• Transcribe audio using OpenAI's Whisper model
• Display transcriptions with pagination
• Sanitize filenames for compatibility
• Support for CUDA acceleration (if available)

Demo

Check out the demo of the application: OpenAI Whisper Transcribe YouTube Videos

Requirements

• Python 3.9+
• PyTorch
• yt-dlp
• whisper
• tiktoken
• streamlit
• FFmpeg (for audio conversion)

Installation

1. Clone this repository:

   git clone https://github.com/RiteshGenAI/openai_whisper_transcribe_yt_videos.git
   cd openai_whisper_transcribe_yt_videos
   

2. Install the required packages:

   pip install -r requirements.txt
   

3. Install FFmpeg if it's not already on your system. Installation methods vary by operating system.

Usage

1. Run the Streamlit app:

   streamlit run .srcapp.py
   

2. Enter a YouTube video URL in the provided input field.

3. The app will download the audio, transcribe it, and display the transcription with pagination.

How it Works

1. Download Audio: The download_audio function uses yt-dlp to download the audio from the provided YouTube URL. It saves the audio as a WAV file.

2. Transcribe Audio: The transcribe_audio function uses OpenAI's Whisper model to transcribe the downloaded audio file.

3. Display Transcript: The display_transcript_with_pagination function splits the transcript into pages and displays them using Streamlit's UI components.

4. Process Audio: The process_audio function orchestrates the entire process, from downloading to transcribing and displaying the result.

Customization

• You can change the Whisper model size by modifying the model_name parameter in the transcribe_audio function.
• Adjust the tokens_per_page parameter in display_transcript_with_pagination to change the amount of text displayed per page.

Note

This application requires a significant amount of computational resources, especially for longer videos. Using a CUDA-enabled GPU can significantly speed up the transcription process.

License

MIT License