Use to generate AI audio from approved lyrics and a style description. Prepares the generation prompt, waits for user to drop in the audio file, then validates it.
Composes a generation prompt from lyrics and style, presents it to the user, then validates the audio file once the user places it.
Current music generation models (Suno, Udio, etc.) are not on Replicate. This skill is model-agnostic — the agent prepares the prompt and the user generates externally.
<chapter-dir>/lyrics.txt — from generating-lyrics<chapter-dir>/lyrics.txtLyrics:
<full lyrics text>
Style: <genre>, <BPM> BPM, <instrumentation>, <mood>
Reference: <URL if provided>
<chapter-dir>/song.mp3source .env && node skills/generating-song/scripts/validate-song.js \
--file <chapter-dir>/song.mp3
aligning-lyrics<chapter-dir>/song.mp3 (placed by user; validated by script)source .env && node skills/generating-song/scripts/validate-song.js --file <path>
Exits 0 if file exists, has a supported extension (.mp3 .wav .flac .m4a .ogg), and size > 0. Exits 1 with error to stderr otherwise.
Transition to aligning-lyrics.
Log to {project_dir}/logs/production.jsonl. See skills/lib/logging-guide.md for schema.
On invocation — key inputs: lyrics_path, style_description
On completion — key outputs: audio_path, duration_s