---

name: openrouter-text2music description: Generate atmospheric sound-design / SFX (NOT speech, NOT the front-of-blog BGM) via OpenRouter using Google Lyria 3 Pro.

openrouter-text2music

Text → audio via OpenRouter. Default model: google/lyria-3-pro-preview.

Role in Data2Story: atmospheric sound-design / SFX only — never the front BGM. Use this for a sound the story needs that has NO findable real recording: an ambient bed (tension drone, texture, room tone) or best-effort sharp foley (e.g. a glass-shatter texture). Lyria's strength is ambient/textural beds, not crisp transients, so sharp foley is best-effort. The result is an opt-in, event/section-bound, reader-controlled element. The front-of-blog BGM must be a SOURCED real track (sourced_bgm via the Scout) — never generated here.

⚠️ This is a generative audio model, not TTS. It produces 48kHz stereo audio; for narration/voiceover use a dedicated TTS tool (e.g., OpenAI tts-1, ElevenLabs).

Usage

Resolve TOOL_DIR = the directory containing this SKILL.md. Commands below use TOOL_DIR as a symbolic placeholder; replace it with the resolved, quoted path before running Bash.

export OPENROUTER_API_KEY=sk-or-v1-...

python3 TOOL_DIR/scripts/generate_music.py \
  --prompt "Low rumbling tension drone, sub-bass swell, no melody, slowly building unease" \
  --download PROJECT_DIR/assets/sfx_drone.wav

SFX prompting

Describe a TEXTURE, not a song — no genre/BPM/lead/vocals. Examples:

• "low rumbling tension drone, sub-bass swell, no melody"
• "sharp glass-shatter texture, brittle high transients" (foley — best-effort)
• "distant wind and creaking-ice ambience, cold and hollow"

Flags

Flag	Default	Description
--prompt	required	Text prompt — describe the atmospheric texture / SFX (mood, materials, motion); avoid song structure
--download	required	Output audio file path
--model	google/lyria-3-pro-preview	Alt: google/lyria-3-clip-preview (shorter clips)

Pricing

• lyria-3-pro-preview: $0.08 per call

Notes

• Request uses POST /api/v1/chat/completions with modalities: ["audio","text"].
• Response parsing handles several shapes: message.audio, a content part with type=audio, or a data:audio/... URL embedded in the text content.
• Output format is typically WAV (48kHz stereo); the script saves whatever bytes the API returns — choose the extension to match.