mist-voicelisted

# /mist-voice MIST's actual voice (Thomasin McKenzie's MIST from *Pantheon*), cloned offline with XTTS-v2 and running entirely on this Mac. No cloud TTS, ever. **The one number that governs every decision here:** synthesis is *slower than real time* (RTF ~1.78 CPU, ~1.58 MPS). A sentence takes 6-8s to generate. So this toolset is for **pre-rendered** audio, never live conversation. | Alex wants | Mode | Section | |---|---|---| | This reply, spoken, in the chat | **Speak this reply** | §1 | | A note / briefing / doc read aloud to a file | **Narrate** | §2 | | One line spoken out of the speakers | **Say** | §3 | | Audio turned into text | **Transcribe** | §4 | | The voice itself changed (accent, clips, model) | **Voice config** | §7 | --- ## §1 Mode: Speak this reply The headline mode. Alex asks for "an audio response" / "say that out loud" / "audio version" and gets a playable track inline alongside the written reply. **Standing rule: the written reply always ships too.** The audio is an addition, never a replacement. He shouldn't have to press play to find out what was said. ```bash cd "/Users/alexhedtke/Documents/Exobrain harness" printf '%s\n' "<the spoken text>" | mist-voice/.venv/bin/python \ mist-voice/scripts/narrate.py - -o "tmp/audio/<slug>.mp3" ``` Then embed it in the reply, path **raw with literal spaces**: ```  ``` Rules for this mode: - **Check the service first** (§5). Cold,