Song Generator GPT Spec

A specification for using GPT and AI tools to generate emotionally resonant music. From prompt to production.

The Core Workflow

AI does not replace the composer. AI extends the composer's range. This spec is a bridge between human intention and machine generation.

Phase 1: Emotional Targeting

Before generating anything, define the emotional target. Use the 27-emotion framework from the All Emotions Album.

Target emotion: Nostalgia Intensity: 7/10 Context: Late summer evening, transition between chapters Personal anchor: The smell of rain on hot pavement

Phase 2: Parameter Extraction

Translate the emotional target into musical parameters.

Tempo

Nostalgia: 60-80 BPM. Slower than heartbeat. Time stretches.

Key

Major with modal mixture. Mixolydian or Lydian for golden-hour quality.

Texture

Warm pads, distant piano, subtle vinyl crackle. Layered but not dense.

Dynamics

Soft attack, slow swell, gentle decay. No sudden changes.

Timbre

Electric piano, nylon guitar, warm synth pads. Avoid bright or harsh tones.

Phase 3: Prompt Engineering

The prompt is the composition. Craft it with precision.

Generate a 32-bar chord progression in D Mixolydian, 72 BPM, for a nostalgic late-summer evening scene. Include: - Warm electric piano voicings - Suspended chords for openness - A brief modulation to the relative minor for contrast - Return to tonic for resolution Emotional target: Bittersweet nostalgia. Not sad. Not happy. Both.

Phase 4: Generation Tools

Phase 5: Human Curation

AI generates options. Human selects. Human edits. Human decides when it is done. The spec emphasizes curation over generation.

$0

The spec is free. The tools are available. The music is yours.

Get the All Emotions Framework → Browse All Products
💰 Support Drake Enterprise — 28 books & protocols, instant download Browse Library →