alphacep/kaldi-vosk-server-gpu repository overview

This is a core image suitable to mount different languages.

A websocket server based on Kaldi and Vosk library with English model included. Supports fast decoding on GPU card.

For details see https://github.com/alphacep/vosk-server⁠

Usage:

Start the server

docker run --runtime=nvidia -v model:/opt/vosk-server/websocket-gpu-batch/model -p 2700:2700 alphacep/kaldi-en-gpu
cd /opt/vosk-server/websocket-gpu-batch
python3 ./asr_server_gpu.py

Check the sample code from https://github.com/alphacep/vosk-server/blob/master/websocket/test.py⁠, run

./test.py test16k.wav

You can try with any wav file which has proper format - 16khz 16bit mono PCM. Other formats has to be converted before decoding.

For client samples in other programming languages see https://github.com/alphacep/vosk-server/tree/master/client-samples⁠