Yt2txt

Description

This project allows you to download YouTube videos, extract audio, and transcribe the audio files into text using the Faster Whisper library.

Installation

1. Prerequisites

• Python 3.8 or newer must be installed. Download Python.
• A GPU compatible with CUDA (e.g., NVIDIA RTX 4090) is recommended for better performance on Windows/Linux.
• macOS users with Apple Silicon (M1/M2/M4) can use the CPU version optimized for Metal Performance Shaders (MPS).

2. Set Up a Virtual Environment

Create and activate a virtual environment:

On Windows

python -m venv .venv
.\.venv\Scripts\Activate

On macOS/Linux

python3 -m venv .venv
source .venv/bin/activate

3. Install Dependencies

Install the required libraries for your project:

pip install -r requirements.txt

4. Install PyTorch

On Windows/Linux with NVIDIA GPU (CUDA)

Install PyTorch with CUDA support. Replace cu118 with your specific CUDA version if necessary.

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

On macOS with Apple Silicon (MPS)

Install the MPS-optimized version of PyTorch:

pip install torch torchvision torchaudio

For CPU-only

For systems without a GPU:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

5. Verify Installation

Run the following script to ensure PyTorch is configured correctly:

python -c "import torch; print(torch.cuda.is_available()); print(torch.backends.mps.is_available()); print(torch.__version__)"

• On Windows/Linux, torch.cuda.is_available() should return True.
• On macOS with MPS, torch.backends.mps.is_available() should return True.

---

Usage

Run the project with the following command:

python your_script.py --url "https://youtube.com/..." -o output -m large-v3

• --url: The URL of the YouTube video or playlist.
• -o: The output directory for the transcribed files.
• -m: Whisper model size to use (e.g., large-v3).

---

Additional Notes

Installing CUDA for Windows/Linux with GPU

If using CUDA on an NVIDIA GPU, make sure to:

• Install and update the NVIDIA drivers.
• Optionally install the CUDA Toolkit. Download from NVIDIA CUDA Toolkit.

Optimization for macOS

PyTorch leverages MPS (Metal Performance Shaders) to accelerate computations on Apple Silicon. No additional configuration is required.

---

Development and Contributions

If you'd like to contribute, ensure you test the project on multiple platforms and configurations (Windows, macOS, GPU, CPU).

---

Authors

Project developed by [Your Name/Team].