qhduan / streaming-sensevoice

Pseudo Streaming SenseVoice with Hotwords

streaming-sensevoice

Streaming SenseVoice processes inference in chunks of SenseVoice.

Usage

• transcribe wav file

$ python main.py

• transcribe from microphone

$ python realtime.py

• transcribe from websocket

A basic WebSocket service built with Recorder and FastAPI; the frontend uses MP3 format to transmit audio information to reduce latency and increase stability.

pip install -r requirements-ws-demo.txt
python realtime_ws_server_demo.py

# check cli options
python realtime_ws_server_demo.py --help