Talk to your Mac. It talks back, in your own cloned voice. Zero internet.
A continuous hands-free voice loop for macOS: Apple's SFSpeechRecognizer for the ears, cloned-voice TTS for the mouth, and nothing in between that touches a network. Built originally to drive claude-code-local, but it stands alone.
What's in it
- Continuous listening — no wake word gymnastics, no push-to-talk
- On-device speech recognition via Apple's Speech framework
- Cloned-voice text-to-speech for replies
- One-command launcher (
narrative-claude.sh) - Works on a plane. MIT licensed.
This is the ears + mouth of a local-first ambient-computing stack — pair it with claude-code-local (the brain) and browser-agent (the hands).
Tagging the current state as v1.0.0 so there's something to watch for updates against.