Whisper Server

Whisper Server is a package designed to deploy and utilize the Whisper speech recognition model efficiently. Whisper is a versatile speech recognition model trained on a vast and diverse dataset of audio samples. It boasts the capability to handle various tasks such as multilingual speech recognition, speech translation, and language identification simultaneously.

Features

• State-of-the-Art Speech Recognition: Whisper utilizes advanced machine learning techniques to provide accurate and reliable speech recognition.
• Multilingual Support: With its multitasking capabilities, Whisper can recognize speech in multiple languages.
• Speech Translation: Whisper can translate recognized speech into different languages, facilitating communication across linguistic barriers.
• Language Identification: Identify the language of the spoken input accurately, enabling dynamic language switching.

Installation

To install Whisper Server, follow these steps:

1. Clone the repository from GitHub:

git clone https://github.com/opennlu/whisper-server.git

2. Navigate to the directory:

cd whisper-server

3. Install dependencies:

pip install -r requirements.txt

Usage

To start Whisper Server from the command line, execute the following command:

python whisper_server.py

API Documentation

Process Audio File

Endpoint: POST http://localhost:28466/whisper

This endpoint allows users to submit an audio file for processing. The audio file should be sent as form-data in the request body.

Request Parameters:

• file: The audio file to be processed. Required.
• language: The returning language to be used. Optional.

curl -X POST \
  -F "file=@/path/to/your/audio/file.mp3" \
  http://localhost:28466/whisper

Docker

Caution

This is experimental, as it will always pull the model files when using docker run, as there is still no volume mount.

CPU only

docker run -d -p 28466:28466 --name whisper-server opennlu/whisper-server

Nvidia GPU

1. Install the Nvidia container toolkit.
2. Run Whisper Server inside a Docker container

docker run -d --gpus=all -p 28466:28466 --name whisper-server opennlu/whisper-server

Contribution

Contributions to Whisper Server are welcome! If you find any bugs or have suggestions for improvements, please open an issue or submit a pull request on GitHub.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

• The Whisper team for developing and open-sourcing this powerful speech recognition model.
• Contributors who help improve and maintain Whisper Server.