voiceover-maker

<objective> Convert text to spoken audio (MP3) using a TTS provider. Pick the right provider based on language + voice character needs. Output an MP3 ready to drop into a video editor / podcast track / reel. Distinct from `music-prompt`: - TTS is speech, not music. Different providers (Eleven TTS, OpenAI gpt-4o-mini-tts) vs. Suno/Udio/Lyria. - No genre / meta-tags / two-box workflow. - Output is one MP3 per call, not a song structure. This skill does NOT: - Generate music (use `music-prompt`). - Compose voice + music together (use `audio-mix-maker` — planned, see ROADMAP). - Lip-sync to a video — that requires separate tooling. - Clone voices — Eleven supports voice cloning via their dashboard, not via this skill (consent + ToS concerns). - Transcribe audio (the opposite direction — speech-to-text — is a separate `transcribe` skill, also on the roadmap). - Mix multiple voices in a single MP3 — run the skill multiple times and stitch externally. </objective> ## ROLE Read the script text + optional voice + optional language → pick provider (Eleven for multilingual / long-form / quality voice control; OpenAI gpt-4o-mini-tts for cheap fast English-first TTS) → call the audio modality runner → save MP3. ## PIPELINE 1. **Resolve script source**: - `--prompt "<text>"` — inline script - `--prompt-file <path>` — read script from file - Or pipe via stdin (the runner accepts that too) 2. **Pick provider** — see `references/voice-picker.md`: - `--model auto`: - Mu...