text-to-speech

Converts a text document to an MP3 audio file via a local TTS pipeline. The skill ships a wrapper script (synth-audio.sh) with a clean text → audio-file interface. The current backend is Kokoro-82M; swapping backends is a config change.

Does NOT install any TTS backend. See Backend Dependencies.

v2.0.0 breaking change: the narrative rewrite (L1) no longer runs automatically inside synth-audio.sh. If you're a driving Claude Code agent, dispatch a rewrite subagent first — see L1 Narrative Rewrite below — this is the normal way to use the skill now, not an edge case. A bare synth-audio.sh input.md output.mp3 with no existing narrative.txt now fails loudly instead of silently shelling out to claude --print (that nested call could inherit your session's output style/CLAUDE.md and leak meta-commentary into the audio — see L1 Narrative Rewrite for why). For headless/standalone use, pass --allow-inline-llm-rewrite.

Quick Start