whisperd 🎙️

A simple HTTP server written in Rust for the OpenAI Whisper speech-to-text model.

Features ✨

• 🎧 Transcribe audio files
• 🔄 OpenAI API compatibility
• 🌈 Models
  • tiny.en
  • tiny
  • base.en
  • base
  • small.en
  • small
  • medium.en
  • medium
  • large
  • large-v1
• 🌎 Languages

  • 🇬🇧 English (en)

  • 🇨🇳 Chinese (zh)

  • 🇩🇪 German (de)

  • 🇪🇸 Spanish (es)

  • 🇷🇺 Russian (ru)

Quickstart 🚀

1. Clone this repository:

git clone https://github.com/tiero/whisperd.git

2. Navigate to the repository and build:

cd whisperd
cargo build --release

3. Run the server:

./target/release/whisperd serve --model_path path_to_whisper_model

Now, the server is running at http://localhost:8000 and ready to transcribe!

Usage 🛠️

CLI Commands

• Start the transcription server:

whisperd serve --port 5000 --model_path <path_to_model> 

• Transcribe a given audio file (this downloads the model automatically from HuggingFace):

whisperd transcribe --audio <path_to_audio>

For more advanced options, use:

whisperd --help

Contribution 🤝

Pull requests and issues are welcome!

License 📜

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