Skip to content

Implemented Whisper speech-to-text conversion using Cobra for detecting if someone is speaking.

Sebastian Olsson requested to merge 142-implement-whisper-with-cobra into dev

Whisper works by saving audio message bytes to BytesIO buffer and flushing it when the person has finished speaking.

  • Created Whisper node.
  • Created cobra.py running a ROS service to detect if someone is speaking in given audio data.
  • API keys are retreived by loading a *_cred.json file (both GPT3 and Cobra).
  • Whisper adds the previously received audio bytes when starting to listen, in case of cobra not hearing the first letters/words and to prevent these being cut off.

Closes #142

Merge request reports