- Capture audio input and send to a local voice recognition engine
- Play back the captured sound
- Record voice to a ".wav" file
docker run -d -p 2700:2700 alphacep/kaldi-en:latest
This runs official VOSK-server docker image../vosk-sound-test
It starts capturing and displays the words you say, also saves audio to file out.wav- Press
<Enter>
to stop capturing and play back - Press
<Enter>
again to exit.
- Low sound quality
Maybe you're using some bluetooth airbuds like "airpods". For system like Linux, the input sound frequency is limited to 8000 at bluetooth stack, 16000 is a minimal frequency for VOSK to work well. A dedicated wired/wireless microphone should work.
- It doesn't work at all
Open the system sound manager, verify the recording device while this program is capturing. Sometimes a wrong device is choosed by default.
- Install Golang
git clone https://github.com/aj3423/vosk-sound-test
cd vosk-sound-test
go build .
MIT