I want local speech recognition for Web Speech

I would like to see speech recognition (STT) implemented locally within the browser in an extensible way. This would avoid the necessity of recording the user voice (with or without notice or permission) and sending that personally-identifiable, sensitive, biometric data to a remote web service.

The current workaround is to not use SpeechRecognition in Chrome or Chromium browsers because implementers do not give notice or ask for permission to record and ship the user’s voice to an undisclosed remote web service. A viable alternative not shipped with the browser is PocketSphinx.