Adventures in voice recognition

[qwertie]

So GitHub Copilot now has a voice-to-text extension called "VS Code Speech", which includes an input option that supports push-to-talk (you hold Ctrl+I but it's a little weird because you can't release Ctrl+I until after it has converted your speech to text), and an optional handsfree mode that is activated by saying "hey code". Although the description says "No internet connection is required, the voice audio data is processed locally on your computer", I couldn't find a way to do voice recognition without immediately sending the text to the (online) Copilot AI... which makes it less useful and pretty stupid actually, since the recognized text is rarely perfect and needs corrections, plus there's a much worse UX pitfall I won't get into.

Then I was amazed to discover that Chrome has built-in voice recognition via the Web Speech API (documented on MDN but not available in Firefox). So, one need only visit a demo page like this to get voice recognition. There's also a "Voice Recognition Anywhere" browser extension based on this, which has tons of features ... but seems quite buggy / is confusing to use, so has not convinced me to switch away from Firefox.

So I have three voice recognition tools including whisper-writer, but I can't nominate a "best" one yet.