Let me just repost this tip:

If you add Deepgram listen API compatibility, you can do live transcription via either Deepgram (cloud) or OWhisper (local): https://news.ycombinator.com/item?id=44901853