whisper-streaming 0.1.0

Providing easy-to-use and extensible STT (Speech-To-Text) implementation with Whisper-like ASR (Automatic Speech Recognition) models.

whisper-streaming

Providing easy-to-use and extensible STT (Speech-To-Text) implementation with Whisper-like ASR ( Automatic Speech Recognition) models.

[!WARNING]
This project is currently in Alpha State. It is probably not stable over a long time and can have unexpected errors. You could help to push this project out of Alpha and Beta state by testing and reviewing it. Thanks!

Index

• Appreciation
• Components
  • Backend
  • Receiver
  • Sender
• Installation
• Development
  • Installation of Prerequisites
  • Build executables
• Documentation
  • Installation of Prerequisites
  • Build
• License
• Appendix
  • Python venv
  • Links

Appreciation

This project is the result of a rework of the ideas and prototype implementation created by Dominik Macháček, Raj Dabre, Ondřej Bojar (Original Repository). It is neither official nor fully API- and function-compatible with its original implementation.

Please have a look at their publication: ACL Anthology Bibtex citation

@inproceedings{machacek-etal-2023-turning,
    title = "Turning Whisper into Real-Time Transcription System",
    author = "Mach{\'a}{\v{c}}ek, Dominik  and
      Dabre, Raj  and
      Bojar, Ond{\v{r}}ej",
    editor = "Saha, Sriparna  and
      Sujaini, Herry",
    booktitle = "Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics: System Demonstrations",
    month = nov,
    year = "2023",
    address = "Bali, Indonesia",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.ijcnlp-demo.3",
    pages = "17--24",
}

With the usage of this project, you agree to the license terms, found in the License chapter. This project is not affiliated with the paper, original implementation or their authors. It is just reimplementing their ideas in a more modern und easier to use and adapt way, respecting the license agreement.

Components

Backend

Currently following backends are implemented:

• Faster-Whisper

Receiver

"Receiver" are mechanisms to input data into the ASR model. Out-of-the-box support for:

• ALSA (Advanced Linux Sound Architecture)
• File - Audio file

Sender

"Sender" are mechanisms to output data out of the ASR mode. Out-of-the-box support for:

• Print - Simple console output via "print"
• WebSocket (Client) - Output via network protocol

Installation

This library can be easily installed with pip:

pip install whisper-streaming

The integration of different backends are installed via following extras:

• [all] - installs all backends
• [faster-whisper] - Faster-Whisper

Development

Installation of Prerequisites

• Python 3 (latest) (https://www.python.org/downloads)
• Python venv (optional, recommended) _{GoTo Installation}
• Installation of development requirements:

pip install -r requirements/dev/requirements.txt

• Installation of requirements:

pip install -r requirements/library/requirements.txt

• backend requirements:

pip install -r requirements/library/requirements_faster_whisper.txt

Build executables

python3 -m build

Documentation

Installation of Prerequisites

• Python 3 (latest) (https://www.python.org/downloads)
• Python venv (optional, recommended) _{GoTo Installation}
• Installation of requirements:

pip install -r requirements/docs/requirements.txt

Build

rm -rf docs/_build
sphinx-build -M html docs/ docs/_build

If the developer documentation should be built, the following script can be used:

rm -rf docs/_build
sphinx-build -M html docs/ docs/_build -t Internal

License

This project is published under Apache License, Version 2.0 - please comply with it, if you use/modify/distribute it. The license can be found in "LICENSE". The original implementation is published under MIT and is mentioned at places in this project where it still applies. The license can be found in "LICENSE-MIT". You have to distribute at least these both licenses - in addition to your compliant license.

Appendix

Python venv

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip setuptools

Links

• Project Home
• Original Implementation
• Original Paper