YouTube Video Summarizer using Ollama or Gemini

This Python script extracts subtitles from a YouTube video or an entire playlist, sends the transcript to either a locally running Ollama instance or the Google Gemini API, generates a detailed list of key points, and saves the results to a text file.

Features

• Processes single YouTube videos or entire playlists.
• Automatically fetches available subtitles (prioritizes languages set in config or command line).
• Cleans subtitle text by removing timestamps and annotations.
• Supports two summarization engines:
  • Ollama: Uses a local Ollama instance.
  • Gemini: Uses the Google Gemini API (requires API key).
• Configuration managed via config.yml (API keys, model preferences, URLs, etc.).
• Command-line arguments to override configuration and select engine.
• Appends results (Title, URL, Detailed Points) to a specified output file.
• Includes error handling and progress reporting.
• Configurable delay between processing videos in a playlist.

Prerequisites

• Python 3.x: Ensure you have Python installed.
• Ollama (Optional): If using the ollama engine, you need a running Ollama instance accessible from where you run the script. Download and install it from https://ollama.com/. Make sure the desired model is pulled (e.g., ollama pull llama3:latest).
• Google Gemini API Key (Optional): If using the gemini engine, you need an API key from Google AI Studio (https://aistudio.google.com/app/apikey).
• Required Python Libraries: Install the necessary libraries using pip.

Installation

1. Clone the repository (or download the script):

   # If you have a git repo setup
   # git clone <your-repo-url>
   # cd <your-repo-directory>
   
   # Or just ensure you have the script (e.g., app.py) and create config.yml

2. Create requirements.txt: Create a file named requirements.txt with the following content:

   requests
   pytubefix
   google-generativeai
   PyYAML

3. Install dependencies:

   pip install -r requirements.txt

4. Create config.yml: Create a file named config.yml in the same directory as the script. You can copy and paste the example below and modify it with your settings.

   # config.yml Example
   
   # Default summarization engine ('ollama' or 'gemini')
   engine: ollama
   
   # Ollama settings (used if engine is 'ollama')
   ollama_url: "http://localhost:11434/api/chat"
   ollama_model: "llama3:latest" # Make sure you have pulled this model
   
   # Gemini settings (used if engine is 'gemini')
   # Get your key from [https://aistudio.google.com/app/apikey](https://aistudio.google.com/app/apikey)
   gemini_api_key: "YOUR_GEMINI_API_KEY_HERE" # IMPORTANT: Keep this secure!
   
   # General settings
   language_preference: ['en', 'a.en'] # Preferred subtitle languages (ISO 639-1 codes)
   output_file: "output.txt"
   playlist_video_delay_seconds: 2 # Delay between videos in a playlist
   

   Important: Replace "YOUR_GEMINI_API_KEY_HERE" with your actual API key if you plan to use the Gemini engine. Consider using environment variables or other secure methods for managing API keys in production environments.

Configuration

The script reads its configuration from config.yml. The key settings are:

• engine: Specifies the default summarization engine (ollama or gemini). Can be overridden with the --engine argument.
• ollama_url: The URL for your local Ollama API endpoint.
• ollama_model: The specific Ollama model to use (ensure it's pulled).
• gemini_api_key: Your Google Gemini API Key. Required if using the gemini engine.
• language_preference: A list of preferred language codes for subtitles (e.g., en for English, es for Spanish, a.en for auto-generated English).
• output_file: The default file where results are saved.
• playlist_video_delay_seconds: Time in seconds to wait between processing videos in a playlist to avoid rate limiting.

Most configuration values can be overridden using command-line arguments.

Usage

Ensure your Ollama instance is running if you intend to use the ollama engine.

Command-line arguments:

• -u URL or --url URL: Process a single YouTube video URL.
• -p URL or --playlist URL: Process all videos in a YouTube playlist URL.
• --engine {ollama,gemini}: Override the summarization engine specified in config.yml.
• -o FILE or --output-file FILE: Specify the output file path (overrides config.yml).
• -m MODEL or --model MODEL: Specify the Ollama model name (overrides config.yml, only applies if engine is ollama).
• --ollama-url URL: Specify the Ollama API endpoint URL (overrides config.yml, only applies if engine is ollama).
• --gemini-api-key KEY: Provide your Gemini API key via command line (overrides config.yml, only applies if engine is gemini). Use quotes if your key contains special characters.
• -l LANG [LANG ...] or --lang LANG [LANG ...]: Specify preferred subtitle language codes (overrides config.yml).

Examples:

1. Process a single video using the default engine (from config.yml):

   python app.py -u "https://www.youtube.com/watch?v=dQw4w9WgXcQ"

2. Process a playlist using the Gemini engine (API key must be in config.yml or provided):

   python app.py --engine gemini -p "https://www.youtube.com/playlist?list=PL..."

3. Process a playlist using Gemini, overriding the API key and output file:

   python app.py --engine gemini --gemini-api-key "AIzaSy..." -p "https://www.youtube.com/playlist?list=PL..." -o gemini_summaries.txt

4. Process a single video using Ollama with a specific model, preferring Spanish subtitles:

   python app.py --engine ollama -m mistral:latest -u "https://www.youtube.com/watch?v=..." -l es en

Output

The script appends the results for each processed video to the specified output file (default: output.txt from config.yml or via -o). Each entry follows this format: