https://github.com/jeanjerome/echoinstone
EchoInStone is an audio processing tool that transcribes, diarizes, and aligns speaker segments from audio files, prioritizing accuracy and reliability.
https://github.com/jeanjerome/echoinstone
alignment diarization localhost pyannote python transcribe whisper
Last synced: about 1 month ago
JSON representation
EchoInStone is an audio processing tool that transcribes, diarizes, and aligns speaker segments from audio files, prioritizing accuracy and reliability.
- Host: GitHub
- URL: https://github.com/jeanjerome/echoinstone
- Owner: jeanjerome
- Created: 2025-02-13T11:40:12.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-03-20T07:03:53.000Z (2 months ago)
- Last Synced: 2025-04-10T00:53:30.292Z (about 1 month ago)
- Topics: alignment, diarization, localhost, pyannote, python, transcribe, whisper
- Language: Python
- Homepage:
- Size: 999 KB
- Stars: 7
- Watchers: 2
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# EchoInStone
**EchoInStone** is a comprehensive audio processing tool designed to transcribe, diarize, and align speaker segments from audio files with a focus on achieving the most accurate and faithful transcription possible. It supports various audio sources, including YouTube videos and podcasts, and provides a flexible pipeline for processing audio data, prioritizing precision and reliability over speed.
## Features
- **Transcription**: Convert audio files into text using state-of-the-art automatic speech recognition (ASR) model, `Whisper Large v3 Turbo`.
- **Diarization**: Identify and separate different speakers in an audio file with the cutting-edge model, `Pyannote Speaker Diarization 3.1`.
- **Alignment**: Align transcribed text with the corresponding audio segments using a customized algorithm tailored to be highly efficient and faithful to the outputs of Whisper and Pyannote, `SpeakerAlignement`.
- **Flexible and Extensible Pipeline**: Easily integrate new models or processing steps into an orchestrated pipeline, `AudioProcessingOrchestrator`.
> Note: The current version of EchoInStone is a preliminary release. Future updates will include more flexible configuration options and enhanced functionality.
## Installation
### Prerequisites
- Python 3.11 or higher
- Poetry (dependency management tool)
### Steps
1. **Clone the repository**:
```bash
git clone https://github.com/jeanjerome/EchoInStone.git
cd EchoInStone
```
2. **Install dependencies using Poetry**:
```bash
poetry install
```
3. **Configure logging** (optional):
- The logging configuration is set up to output logs to both the console and a file (`app.log`). You can modify the logging settings in `logging_config.py`.
4. **Configure Hugging Face Token**:
- Add your Hugging Face token to this file. You can obtain a token by following these steps:
1. Go to [Hugging Face Settings](https://huggingface.co/settings/tokens).
2. Click on "New token".
3. Copy the generated token and paste it into the `EchoInStone/config.py` file as shown below:
```python
# EchoInStone/config.py
# Hugging Face authentication token
HUGGING_FACE_TOKEN = "your_token_here"
```
## Usage
### Basic Example
To transcribe and diarize a YouTube video, you can run the following command:
```bash
poetry run python main.py
```
- ``: The URL of the audio input (YouTube, podcast, or direct audio file).
### Command-Line Arguments
- **`--output_dir`**: Directory to save the output files. Default is `"results"`.
```bash
poetry run python main.py --output_dir
```
- **`--transcription_output`**: Filename for the transcription output. Default is `"speaker_transcriptions.json"`.
```bash
poetry run python main.py --transcription_output
```
### Examples
- **Transcribe and diarize a YouTube video**:
```bash
poetry run python main.py "https://www.youtube.com/watch?v=plZRCMx_Jd8"
```
- **Transcribe and diarize a podcast**:
```bash
poetry run python main.py "https://radiofrance-podcast.net/podcast09/rss_13957.xml"
```
- **Transcribe and diarize a direct MP3 file**:
```bash
poetry run python main.py "https://media.radiofrance-podcast.net/podcast09/25425-13.02.2025-ITEMA_24028677-2025C53905E0006-NET_MFC_D378B90D-D570-44E9-AB5A-F0CC63B05A14-21.mp3"
```
## Testing
To run the tests, use the following command:
```bash
poetry run pytest
```
This command will execute all the tests, including BDD tests, to ensure the functionality of the application.
## Configuration
### Logging
Logging is configured to output messages to both the console and a file (`app.log`). You can adjust the logging level and format in the `logging_config.py` file.
### Models
- **Transcription Model**: The default transcription model is `openai/whisper-large-v3-turbo`. You can change this by modifying the `model_name` parameter in the `WhisperAudioTranscriber` initialization.
- **Diarization Model**: The default diarization model is `pyannote/speaker-diarization-3.1`. You can change this by modifying the model loading code in the `PyannoteDiarizer` class.
## Contributing
Contributions are welcome! Please follow these steps:
1. Fork the repository.
2. Create a new branch (`git checkout -b feature-branch`).
3. Make your changes and commit them (`git commit -am 'Add new feature'`).
4. Push to the branch (`git push origin feature-branch`).
5. Create a new Pull Request.
## License
This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.
## Acknowledgments
- Thanks to the open-source community for the various libraries and models used in this project.
- Special thanks to the contributors and maintainers of the models and tools that make this project possible.
## Contact
For any questions or suggestions, please open an issue.