sonicscribe 1.0.0

🎙️ SonicScribe - Transcribe & Translate audio/video files using Whisper and GPT models.

🎙️ SonicScribe

A powerful CLI tool and Python module for transcribing and translating audio/video files using OpenAI's Whisper and GPT models.

Features

• 🎬 Extract audio from various video and audio formats

• 🔤 Transcribe audio using OpenAI's Whisper API

• 🌐 Translate transcriptions to English or other languages

• 📝 Generate transcript and subtitle files (SRT)

• 🗣️ Support for bilingual subtitles

• 📊 Smart handling of large files by automatic chunking

• ⚡ Interactive language selection with auto-detection support

• 🛠️ Modular design for use as a Python library

• 📈 Progress indicators and detailed logging

---

Installation

Prerequisites

• Python 3.8 or higher

• OpenAI API key

Setup

1. Clone the repository:

   git clone https://github.com/tuklu/SonicScribe.git
   
   cd SonicScribe/SonicScribe
   

2. Create a virtual environment and activate it:

   python -m venv venv
   
   # On Windows
   
   venv\Scripts\activate
   
   # On macOS/Linux
   
   source venv/bin/activate
   

3. Install the required packages:

   pip install -r requirements.txt
   

4. Create a .env file in the root directory with your OpenAI API key:

   OPENAI_API_KEY=your_api_key_here
   

---

Usage

Main Transcription and Translation

The main script provides full functionality for processing audio/video files:

python main.py --input "path/to/your/video.mp4" --translate --output-dir "output/folder"

Options

• --input: Path to input audio/video file (required)

• --translate: Enable translation to English (optional)

• --language: Specify the language of the input file (e.g., en, fr, es). If not provided, auto-detection will be used.

• --output-dir: Directory to save output files (default: output/transcripts)

• --whisper-model: Model to use for transcription (default: whisper-1)

• --gpt-model: Model to use for translation (default: gpt-4o-mini)

• --chunk-size: Size of chunks in MB for large files (default: 20)

• --verbose or -v: Enable verbose logging

• --bilingual: Create bilingual subtitles with both original and translated text

---

Interactive Language Selection

When processing files, SonicScribe provides an interactive language selection feature:

• You can select a language from a predefined list using arrow keys.

• You can manually input a language code by typing /.

• If no language is selected, SonicScribe will auto-detect the language using GPT, with a warning about potential additional API costs.

---

Standalone SRT Translation

If you already have an SRT file and just want to translate it:

python translate_srt.py --input "path/to/your/subtitles.srt" --bilingual

Options

• --input: Path to input SRT file (required)

• --output: Path to output SRT file (default: input_english.srt)

• --model: GPT model to use for translation (default: gpt-4o-mini)

• --bilingual: Create bilingual SRT with original and translated text

• --language: Specify the language of the input subtitles. If not provided, auto-detection will be used.

---

Examples

Basic Transcription

python main.py --input "lecture.mp4"

This will:

• Extract audio from lecture.mp4

• Transcribe the audio using Whisper API

• Save a transcript and SRT file to the default output directory

Transcription with Translation

python main.py --input "foreign_movie.mp4" --translate --gpt-model "gpt-4o"

This will:

• Extract audio from foreign_movie.mp4

• Transcribe the audio using Whisper API

• Translate the transcription to English using GPT-4o

• Save all output files to the default directory

Bilingual Subtitles

python main.py --input "interview.mp3" --translate --bilingual

This will:

• Extract audio from interview.mp3

• Transcribe the audio using Whisper API

• Translate the transcription to English

• Create a bilingual SRT file with both original and translated text

Translating Existing Subtitles

python translate_srt.py --input "movie.srt" --bilingual --model "gpt-4o-mini"

This will:

• Read the existing SRT file

• Translate the subtitles to English using GPT-4o-mini

• Create a bilingual SRT with both original and translated text

---

Output Files

SonicScribe generates several types of output files:

• {filename}_transcribed.txt: Plain text transcript

• {filename}.srt: SRT subtitle file with timestamps

• {filename}_bilingual.srt: Optional bilingual SRT file (when using --bilingual)

• {filename}_english.srt: Translated SRT file (when using translate_srt.py)

---

Handling Large Files

SonicScribe automatically handles large audio files:

• Files smaller than 25MB are processed directly through the Whisper API.

• Larger files are split into chunks, processed separately, and then recombined.

• The --chunk-size parameter controls the size of these chunks (default: 20MB).

---

Troubleshooting

API Key Issues

If you encounter "API key not found" errors:

• Ensure your .env file exists in the project root directory.

• Verify that your API key is correct and active.

• Try setting the API key directly in your environment.

File Format Problems

If SonicScribe fails to process your file:

• Verify the file exists and is not corrupted.

• Check that the format is supported (mp4, mkv, mov, mp3, wav, etc.).

• Try converting the file to a more standard format like MP4 or WAV.

Memory Issues with Large Files

If you encounter memory errors with very large files:

• Try reducing the --chunk-size parameter.

• Ensure your system has sufficient free memory.

• Consider pre-splitting very large files manually.

---

Limitations

• OpenAI API rate limits may affect processing speed.

• Transcription quality depends on audio clarity.

• Translation quality varies by language and content complexity.

• Processing very large files (multiple hours) can take significant time.

---

Logging

SonicScribe logs all operations to the logs directory. If you encounter issues, check the log files for detailed information. Use the --verbose flag for more detailed logging.

---

License

[Your license information here]

---

Support

For issues, questions, or contributions, please create an issue on the GitHub repository.