SubGen-WhisperX

A powerful subtitle generation tool using WhisperX for accurate speech-to-text transcription with precise timestamp alignment.

Features

• 🎯 Accurate speech recognition using WhisperX
• ⚡ GPU acceleration support (CUDA)
• 🎵 Handles both video and audio files
• 📁 Batch processing support
• ⏱️ Performance timing and logging
• 🔧 Multiple language model options
• 🎛️ Configurable compute device selection
• 🌐 Multi-language support with auto-detection
• 📑 Text file input support for batch processing
• 🔄 Automatic model size selection

Prerequisites

• Python 3.10 or later
• NVIDIA GPU with CUDA 12 support (optional, for GPU acceleration)
• Latest driver from NVIDIA (Very important, when using cuda flag)
• FFmpeg
• Git

Installation

The easy way

1. Clone the repository:

git clone https://github.com/protik09/subgen-whisperx.git --depth=1
cd subgen-whisperx

1. Create and activate a virtual environment:

In Powershell:

.\uv_init.ps1

or in bash

.\uv_init.sh

Usage

Basic Usage

Generate subtitles for a single video file:

python subgen_whisperx.py -f path/to/video.mp4

Advanced Options

Process all media files in a directory:

python subgen_whisperx.py -l en -d path/to/directory

Specify compute device and model size:

python subgen_whisperx.py -f video.mp4 -c cuda -m medium

Set logging level:

python subgen_whisperx.py -f video.mp4 -l DEBUG

Command Line Arguments

Argument	Description	Default
-f, --file	Path to input media file	None
-d, --directory	Path to directory containing media files	None
-c, --compute_device	Device for computation (cuda or cpu)	Auto
-m, --model_size	WhisperX model size	Auto
-l, --language	Subtitle language	None
-log, --log-level	Logging level	ERROR
-t-, --txt	Text file with file/folder paths	None

Output

The script generates SRT subtitle files in the same directory as the input media:

• Format: filename.{language}-ai.srt
• Example: Meetings-0822.en-ai.srt for a video called Meetings-0822.mp4

Troubleshooting

If the automatic model selection leads to the CUDA Out of Memory (OOM) Issue, just manually select the next smaller model using the -m flag. For example, if the -m medium flag causes a CUDA OOM, then use -m small.en or -m small.

Performance

• GPU acceleration provides significantly faster processing
• There is CPU fallback if GPU access fails
• Progress indicators show real-time status
• Performance timing information displayed after completion

License

This project operates under the MIT Open Source License

Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Acknowledgments

• WhisperX for the core transcription technology
• FFmpeg for media processing capabilities