Prompt Lab
LLM Concept Prompt
This prompt tells the LLM how to analyze lyrics and what visual concept to generate. It controls the
style direction
(colors, mood, composition).
You analyze song lyrics and extract their emotional and visual essence. Read the lyrics carefully and identify: - The central character or object the song is about - The key visual moment or metaphor (translate abstract emotions into a concrete image) - The emotional tone - The setting or environment implied by the lyrics Return ONLY a JSON object with these fields: - subject: who or what the song is about (specific, concrete — a person, object, or creature) - key_image: one vivid visual moment or metaphor from the lyrics (not abstract — something you can picture) - emotion: 2-3 emotional descriptors - setting: where this takes place (specific environment)
Image Prompt Template
Use
{llm_answer}
placeholder — it will be replaced with the LLM concept fields above. This prompt goes directly to the image model.
Create a square album cover artwork. Song essence: {llm_answer} Visual direction: - Modern illustrated album cover, stylized digital illustration - Cinematic lighting, editorial quality, Spotify/Apple Music aesthetic - All people and objects: illustrated, never photorealistic, never 3D renders - Fill entire canvas edge-to-edge, no borders - Simple scene readable at small thumbnail size - DEEP BLACK background with vivid saturated color accents - High contrast between dark areas and bright elements CRITICAL — No text: Do NOT include any text, letters, numbers, words, symbols, typography, watermarks, signatures, logos The image must be purely visual with zero readable characters
Generate
Song
Baseline
pre-generated
Model: GPT Image 1.5 low ($0.009/img) via fal.ai · LLM: GPT-4.1 Mini via OpenRouter
Copy to editor
