Speech Recognition

This page documents various speech recognition backends supported (or previously supported) by satania-buddy.

Vosk

• This is the default speech recognition backend. It is accurate, very fast, and consumes few resources.

whisper.cpp

• An experimental speech recognition backend, based on whisper.cpp. There’re some disadvantages compared to Vosk
  • It does not support streaming mode, so real-time STT requires some tricks at satania-buddy’s end. The current implementation is not really good and limit the maximum speech buffer to 8 seconds only.
  • It requires a lot of processing power compared to Vosk.

Microsoft Speech Object Library

• A legacy speech recognition backend. Supports Windows only.

CMU Sphinx

• An obsolete speech recognition backend, it was removed in favor of Vosk.