YouTube Video Transcriber & Summarizer

This Python CLI tool allows you to download YouTube videos, transcribe them, and optionally summarize and translate the transcriptions. It supports multiple translation engines, including local and API-based models.

Features

• Download and transcribe YouTube videos using the local Whisper base model
  • Generate SRT files with timestamps
• Translate transcriptions into different languages:
  • Local model: Opus-MT (no key required, runs locally)
  • API-based model: OpenAI's GPT (requires an API key, runs as a service)
• Summarize transcriptions or translated text:
  • Local model: sshleifer/distilbart-cnn-12-6 (no key required, runs locally)
  • API-based model: OpenAI's GPT (requires an API key, runs as a service)
• Interactive wizard: The tool includes a CLI wizard that guides you through the transcription process using click, prompting for input like:
  • The YouTube video URL
  • Whether you want a summary of the transcription
  • Translation options, including language selection and translation model (Opus MT or OpenAI's GPT)
  • If OpenAI's GPT is chosen, the wizard will prompt for your OpenAI API key
• Easy-to-use command-line interface (CLI)

Best Hardware to Run On

To run this tool efficiently, especially for larger videos or when using more resource-intensive models like Whisper and GPT, the following hardware is recommended:

General Recommendations

• GPU: A recent NVIDIA GPU with CUDA support, such as the RTX 30 series or later, significantly accelerates transcription and summarization tasks.
  • For local Whisper-based transcriptions, a GPU will greatly speed up the process.
• CPU: For CPU-only environments, an Intel i7 (10th gen or higher) or AMD Ryzen 7 (4000 series or higher) is recommended for acceptable performance.
• RAM: At least 16 GB of RAM, with 32 GB or more preferred for handling longer videos and larger model workloads.
• Disk Space: Ensure you have enough free disk space to handle video downloads, transcription outputs, and temporary audio files. While the extracted audio files often only consume a few MB, larger video files may require more storage, so having a few GB free is recommended.
• ffmpeg: ffmpeg must be installed and accessible via the system's PATH.

Apple Silicon (M1/M2)

• Apple Silicon (M1/M2/M3): Apple Silicon offers great performance even without a dedicated GPU. The Whisper model, in particular, runs efficiently on the Neural Engine.
  • Whisper: Works well on Apple Silicon, especially when using the coreml version optimized for these processors.
  • Translation & Summarization: OpenAI API-based models (like GPT) run remotely, so performance is not hardware-dependent. For local translation (Opus MT) and summarization (DistilBART), the M1/M2 chip provides excellent performance for most tasks.
• RAM: 16 GB unified memory is generally sufficient for typical use, but for large videos or processing many tasks simultaneously, 32 GB is preferred.
• Disk Space: Similar to other setups, ensure a few GB of free space for video downloads and outputs.
• ffmpeg: Ensure ffmpeg is installed and accessible via PATH, as it's required for video processing.

Installation

Prerequisites

• Python 3.7+
• ffmpeg (Make sure ffmpeg is installed and accessible via your PATH)

Installation

1. Clone the repository:

git clone https://github.com/yourgithubrepo/youtube-transcriber-summarizer.git
cd youtube-transcriber-summarizer

2. Install the package:

python setup.py install

This will install all the required dependencies and make the CLI tool available for use.

Usage

Once installed, you can use the CLI tool like this:

youtube-transcriber transcribe-and-summarize [OPTIONS] VIDEO_URL

Options:

• --summarize: Summarize the transcription.
• --translate TEXT: Translate the transcription to the specified language (e.g., 'es' for Spanish).
• --translator [opus|chatgpt]: Choose the translator: 'opus' for Opus MT, 'chatgpt' for OpenAI GPT.
• --openai-api-key TEXT: Your OpenAI API key (required if using chatgpt).

Example:

youtube-transcriber transcribe-and-summarize https://www.youtube.com/watch?v=example --summarize --translate es --translator opus

This will download the YouTube video, transcribe it, summarize the transcription, and translate it to Spanish using the Opus MT model.