/music-analyze — Full Music Analysis
Analyze a local audio file and output a comprehensive structured JSON containing rhythm, emotion, timbre, tonality, lyrics, onset, and color palette data.
Usage
/music-analyze <audio_file_path>
Accepted Formats
MP3, WAV, FLAC, OGG, M4A, AAC, WMA
Steps
- Validate the audio file path exists and is a supported format
- Run the full analysis pipeline:
python3 -m music_analyzer analyze "<audio_file_path>"
-
Parse the JSON output and present a structured summary to the user:
- Rhythm: BPM, time signature, song structure sections
- Emotion: Primary mood, energy level, valence, genre
- Timbre: Brightness, warmth, dynamic range, MFCC summary
- Tonality: Key, mode, chord progression highlights
- Lyrics: Whether vocals detected, language, transcription preview
- Onsets: Onset rate (for visual sync reference)
- Color Palette: Suggested colors based on mood
-
If the user wants to generate Dreamina prompts or storyboards, suggest:
/music-to-dreamina for Dreamina image/video generation prompts
/music-to-storyboard for shot-by-shot storyboard
/music-color-palette for detailed color scheme
Options
- Add
--no-cache to force re-analysis (skip cached results)
- Add
--no-separation to skip Demucs source separation (faster)
- Add
--output <path> to save JSON to a specific file
Output
The analysis JSON follows the MusicAnalysisResult schema and can be saved and reused as input for the formatter commands (dreamina, storyboard, color-palette).
Error Handling
- If librosa is not installed, instruct user:
pip install -e ~/.claude/plugins/music-analyzer/src/
- If optional features are missing, the tool will degrade gracefully and note which tier is active (lite/standard/full)