← ClaudeAtlas

music-creationlisted

Interview-style music generation. Asks the user about genre, vocals, lyrics, length, optional reference track. Dispatches via ai-asset-generation skill which routes to local-music (default), or elevenlabs / fal-ai on explicit request.
Nagellabs/libi · ★ 2 · AI & Automation · score 75
Install: claude install-skill Nagellabs/libi
# Music Creation > **Related:** when the user supplies a reference track, run > `libi.music_profile({ fileId })` first to seed the answers — it > returns a `suggestedPrompt`, `keyEstimate`, and `descriptors[]` you > can paraphrase in Stage 1. Walk the user through the music spec one question at a time. Don't batch — they want to feel heard. Each answer goes into a running "music brief" you assemble in Stage 8. ## Stage 0 — Frame Ask: "Is this background music for a video, a standalone song, a short jingle, or a beat / instrumental loop?" The answer changes default length + whether vocals are even on the table (a jingle: maybe; a background score: usually not). ## Stage 0.5 — Recreating a video that already has music? (reuse vs generate — ASK FIRST) If you arrived here from a **recreate / mimic** flow (or any flow where a **source video already carries a music bed**), the music is a building block with a source-vs-generate decision — and that decision is the user's, BEFORE the interview below. Do NOT silently generate a new track when the source already has one the user may want to keep. Present the two paths and ask which they want: - **Reuse the original track (recommended when faithfulness matters).** Extract the source's audio and lay it under the new visuals: 1. `libi.extract_audio({ fileId: <sourceVideoId> })` → an audio file (wav/m4a). 2. `libi.audio_add_clip({ pieceId, fileId: <extractedAudioId>, kind: "standalone", startSeconds: 0 })` to place it un