In my last post, I explained how you can easily build your own voice assistant. However, I noticed that my speech was often not understood, or words were recognised that I hadn’t said. At first, I thought it was due to the microphone on the M5 Stack Atom Echo*, but no. The Whisper integration is simply rubbish!
Instead, I now use the OpenAI STT integration, which is worlds better.
To do this, install this integration and create an API key at OpenAI.
Add the integration code to the configuration.yaml file and add your API key.
Now navigate to ‘Settings – Voice assistants – Add assistant’ in Home Assistant.
Give the assistant a name. For ‘speech-to-text’, select ‘openai_stt’.
Pro tip: Click on the three dots in the wizard and select ‘Debug’. There you can see what was received as STT and what was responded to.
