PROMPTS
The Nano Banana Prompt: Image-to-JSON for AI Image Generation ⚡
Convert any image into a machine-readable JSON file that captures every visual detail. Perfect for recreating images in Nano Banana, Flux, DALL·E, and other.
This prompt converts any image into an exhaustive, machine-readable JSON file that captures every discernible visual element, designed to recreate images with maximum fidelity inside generative-image engines like Nano Banana, Flux, DALL·E, and more.
What It Does
Feed this prompt an image, and it will produce a structured JSON that encodes:
- Every visible object: position, dimensions, material, surface properties, color hex codes, lighting behavior
- Background details: texture, patterns, lighting, depth of field
- Composition data: symmetry type, rule of thirds alignment, spatial relationships
- Text content: OCR extraction, font styles, alignment, color
- Lighting analysis: shadow casting, highlight zones, translucency
Why It Matters
Most image-to-prompt workflows lose critical details. This prompt forces a three-pass analysis:
- Macro Sweep: Identify all major elements and overall composition
- Micro Sweep: Scan for fine details, textures, micro-patterns
- Relationship Sweep: Map connections between objects (e.g., "object A resting on object B")
The result is a JSON so detailed that generative engines can reproduce the image with near-perfect accuracy.
How to Use It
- Copy the full prompt from the PDF
- Paste it into ChatGPT, Claude, or Gemini along with your image
- Get your JSON output: a complete, structured breakdown of every visual element
- Feed the JSON into your image generator: Nano Banana, Flux, DALL·E, Midjourney, etc.
Best For
- Recreating reference images in different AI art engines
- Building consistent visual styles across multiple generations
- Creating detailed image specifications for design teams
- Archiving visual compositions in machine-readable format
Download the Full Prompt
The complete prompt with all analysis protocols, object schemas, and output formatting is available in the PDF below.
The real unlock of AI agents is not full autonomy, but human-in-the-loop systems design. Structure your agent inputs with explicit constraints, negative prompts, and automated test checkpoints. When building tools, keep token consumption lean by caching system prompts and isolating tasks into specialized sub-agents.