sense-voice-streaming-asr 0.1.1

Real-time streaming automatic speech recognition (ASR) with support for Chinese, English, Cantonese, Japanese, and Korean languages using SenseVoiceSmall model.

🎙️ SenseVoice Streaming ASR

A lightweight, real-time streaming speech recognition engine powered by SenseVoiceSmall.

✨ Features

• Streaming ASR: lightweight, real-time streaming speech recognition engine, runs on CPU with realtime recongnition.
• Integrated VAD: Built-in Voice Activity Detection (VAD) to detect speech segments.
• Multilingual: Supports Chinese, English, Japanese, Korean, Cantonese out of the box.
• Self-contained: Models bundled inside the package — no external downloads needed.
• Pure Python + ONNX: No heavy dependencies; runs on CPU.

🚀 Installation

pip install sense-voice-streaming-asr

Requires Python ≥ 3.8.

🛠️ Development

To install in development mode:

git clone https://github.com/yourname/sense-voice-streaming-asr.git
cd sense-voice-streaming-asr
git submodule init # for SenseVoice models
pip install -e .

📄 License

Apache 2.0 License.

This project incorporates code from SenseVoice which is licensed under Apache 2.0. The project as a whole is therefore distributed under the Apache 2.0 license. See the LICENSE file for full license text.

🙏 Acknowledgements

• SenseVoice by FunAudioLLM
• SenseVoice ONNX Models
• FSMN VAD
• Kaldi-native-fbank
• ONNX Runtime