Audio Transcription with Whisper

This tool provides an easy way to transcribe or translate audio files using OpenAI's Whisper model. It supports various audio formats and provides optimized performance for different hardware configurations, including Mac M-series processors.

Prerequisites

• Python 3.12
• ffmpeg (required for audio conversion)

Installation

1. Clone the repository:

git clone [repository-url]
cd [repository-name]

2. Install the required packages:

pip install -r requirements.txt

Usage

Basic command structure:

python main.py --input <audio-file> --language <language-code> --task <task-type>

Required Parameters

• --input: Path to the input audio file
• --task: Type of task to perform (transcribe or translate)

Optional Parameters

• --device: Computing device to use (cpu, cuda, mps, or auto [default])
• --language: Language code of the audio (default: "en")
• --model: Model name or path (default: "openai/whisper-large-v3")
• --output: Custom output file path

Examples

1. Basic transcription of an English audio file:

python main.py --input data/audio.mp3 --language en --task transcribe

2. Translation to English:

python main.py --input data/audio.mp3 --language fr --task translate

3. Specifying a device (for Mac M-series processors):

python main.py --input data/audio.mp3 --language en --task transcribe --device mps

Supported Audio Formats

• MP3
• WAV
• M4A (automatically converted to MP3 using FFmpeg)
• MOV (automatically converted to MP3 using FFmpeg)

Output

The tool generates two output files:

1. A VTT file containing timestamped transcriptions
2. A plain text file with the complete transcription

Output files are saved in the same directory as the input file by default.

Performance Notes

• For Mac users with M-series processors, using the mps device provides significant speed improvements compared to CPU processing
• For Windows/Linux users with NVIDIA GPUs, the cuda device will be automatically selected when available
• The tool processes audio in chunks of 30 seconds by default for optimal memory usage

Troubleshooting

1. If you get FFmpeg-related errors:

   • Ensure FFmpeg is installed on your system
   • On Mac: brew install ffmpeg
   • On Ubuntu: sudo apt-get install ffmpeg
   • On Windows: Download from the official FFmpeg website

2. For Mac M-series users:

   • If you experience issues with MPS device, try falling back to CPU with --device cpu