Fails to send microphone input to OpenAI API -- Error in console: "Content-Security-Policy: The page's settings blocked the loading of a resource at https://api.openai.com/v1/audio/transcriptions ("connect-src")." Hope this can be fixed.

Also would be nice to see the use of the OpenAI TTS for speaking the output.

It used to work very well, but it no longer does.

Works pretty good for me, took a minute to set it up, you need to have a paid OpenAI account for it to work and you need to make an API and give the site/add-on various permissions.
After that it works great, the speech-to-text was surprisingly accurate even with my poor enunciation, the text-to-speech is an annoyingly slow speaker but maybe that's normal for TTS, I have no experience with other TTS systems. I wish it could be sped up a bit.
I also wish that when it comes to ChatGTP it would be possible to end messages early instead of having to listen to the entire thing with seemingly no way of aborting.
I did tell the AI to limit it's responses to <50 words which helped.