I added local TTS and STT to apple-llm using what's already on your Mac. 30 minutes of narration renders in under 20 seconds on an M4 Air
I posted apple-llm here a while back: a Node and Python package that calls the LLMs Apple ships with macOS, with no API key. Version 0.3 adds the rest of what's already on your Mac, the system voices and Apple's speech recognition. There's nothing to sign up for, and your audio stays on the machine.