do-voice-recordinglisted

# /do-voice-recording — Text → Spoken Audio The single, simple text-to-speech surface for the whole system. Hand it text, get back an OGG/Opus file. Works from **any** directory on **any** machine, and every other skill that needs synthesis (`/do-presentation` voiceovers, `/do-debrief` voice notes) defers to this rather than reimplementing it. ## Why this skill exists `valor-tts` is registered in `~/src/ai/pyproject.toml [project.scripts]` and is only on `PATH` when that venv is active. From any other project the capability is invisible — which is the whole reason this skill exists: it makes TTS **discoverable and invocable regardless of cwd**, and it is the one place the synthesis mechanics (binary resolution, flags, prosody) are documented. It is listed in every machine's skill registry once `/update` hardlinks `.claude/skills-global/` into `~/.claude/skills/`. ## Resolve the binary (portable) Never assume `valor-tts` is on `PATH`. Resolve it in this order: ```bash TTS="$(command -v valor-tts || true)" [ -z "$TTS" ] && [ -x "$HOME/src/ai/.venv/bin/valor-tts" ] && TTS="$HOME/src/ai/.venv/bin/valor-tts" [ -z "$TTS" ] && { echo "valor-tts not found — is the ~/src/ai venv installed? Run /update there." >&2; exit 1; } ``` `$HOME/src/ai/.venv/bin/valor-tts` is correct on every machine (the repo lives at `~/src/ai` regardless of the OS username). Invoking by absolute path works from any cwd — the binary's shebang points at its own venv interpreter. ## Synthesize ```bash O