AtoS (Audio to Subtitles)

A command-line tool that transcribes audio files using OpenAI's Whisper model.

Features

• Batch process multiple audio files
• Support for multiple languages
• Customizable input and output directories
• Progress tracking and detailed logging
• Summary report after completion

Prerequisites

• Bun runtime
• OpenAI API key

Installation

1. Clone the repository

2. Install dependencies:

   bun install

3. Set up your OpenAI API key:

   • Copy .env.example to create a new .env file
   • Add your OpenAI API key to the .env file:

   OPENAI_API_KEY=your-api-key-here
   

Usage

Basic usage with default options:

bun transcribe

This command will:

• Rename audio files to remove spaces and special characters
• Transcribe all audio files to text
• Format the output with proper line breaks

Additional Commands

• bun rename - Clean up audio file names
• bun format - Format existing transcript files
• bun clean - Remove all files from audio and transcripts directories
• bun youtube - Download audio from YouTube to the ./audio directory

Output

Transcribed files will be saved in the transcripts directory with the same base name as the input audio file.

Notes

• Supports MP3 and WAV audio files
• Optimized for Polish language transcription
• Automatically formats output with proper sentence breaks

Options

• -l, --lang <language> - Specify the language to transcribe to (default: "en")
• -i, --input-dir <directory> - Set input directory (default: "./audio")
• -o, --output-dir <directory> - Set output directory (default: "./transcripts")

Example with options:

bun transcribe --lang pl --input-dir ./my-audio --output-dir ./my-transcripts

Directory Structure

.
├── audio/          # Default input directory for audio files
├── transcripts/    # Default output directory for transcriptions
├── api/            # API related files
├── index.ts        # Main application file
└── utils.ts        # Utility functions

Output Format

Transcriptions are saved as text files in the output directory. The filename will match the input audio file's name (without the audio extension).

Error Handling

• The tool provides detailed error messages for failed transcriptions
• A summary report shows successful and failed transcriptions
• Failed transcriptions don't stop the batch process

Contributing

Feel free to submit issues and pull requests.

License

MIT