This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
Current speech recognition API cannot specify media stream, we could introduce an optional parameter in SpeechRecognition.start() to enable following use cases: 1. multiple microphone selection via gUM media constraints 2. remote audio stream (WebRTC) 3. stream from audio file
We are working on hooking up gUM audio track with WebSpeech in Chrome, but we are not going to support 2, 3 due to concerns of server abuse. That says, we only allow hooking up audio track from microphone to WebSpeech, a track uses non-microphone source (like a file or remote audio track) will be thrown a exception when connecting to WebSpeech. The new API allows WebSpeech to benefit from gUM technologies, ex. AEC. This will substantially improve the recognition performance during a conference call.
This work was started in Chrome in https://crbug.com/408940 but wasn't finished and then removed again. To revisit this issue, I suggest filing an issue on https://github.com/w3c/speech-api and getting implementers talking to each other.