Push-To-Talk Speech Recognition

Description:

This is a demonstration of the Kaldi Speech Recognition Toolkit (https://kaldi-asr.org/) using the Zamia generic English model (https://github.com/gooofy/zamia-speech).

Image installation procedure

docker build -t ptt .

or

podman build -t ptt .

Container run procedure

docker run -p 80:80 ptt

or

podman run -p 80:80 ptt

Test procedure

1. Browse to: http://localhost/

2. When prompted, allow web page access to the microphone.

3. Click and hold the Push-to-Talk button, and say something into the microphone.

Limitations

Mozilla Firefox does not implement the getUserMedia sampleRate constraint (https://bugzilla.mozilla.org/show_bug.cgi?id=1388586).