I would like to use the microphone from webbrowser and send the audio chunks in realtime to a python backend via websockets. The audio data (float32arrays) are sent, but seem to be different than the one created by pyaudio (although I can't identify the difference as both are always between -1.0 and +1.0) and the backend (faster_whisper) cannot properly/at all translate audio into text. The implementation in python only is working (although I am using the very same logic)
In order to avoid a lengthy entry, I am referring to a respective github issue, where both code blocks (1. not [fully] working javascript frontend/python backend + 2. working python-only script)
https://github.com/SYSTRAN/faster-whisper/issues/671
the number of arrays sent and the values (between -1.0 and +1.0) seem to be correct ... running out of ideas on what is wrong ...