In my experience, OpenAI's Whisper speech recognition is beyond anything currently out there. Likely Github will use it on the backend.