Mastering NanoBanana: How to Structure the Perfect JSON Prompt
Introduction
Every creative platform has its quirks. With some, you throw in free-form text and hope the machine guesses right. NanoBanana is different. It wants your instructions in JSON format. That can feel technical at first, but once you understand the rhythm, it becomes a huge advantage. JSON lets you break your vision into structured pieces: characters, environment, style, lighting, camera, and more.
If you treat JSON as a storyboard instead of a burden, NanoBanana will give you consistency, precision, and the freedom to scale ideas without losing quality. This post will walk through the building blocks of a strong NanoBanana JSON, drawing on best practices and examples you can adapt.
Why JSON Works So Well for NanoBanana
Three big reasons:
- Clarity – Instead of mixing mood, setting, and camera angle in a single sentence, you give NanoBanana labeled fields. There’s no guesswork.
- Consistency – Once you design a schema, you can swap variables (a desert for a forest, a tuxedo for a spacesuit) without rewriting everything.
- Automation – JSON is machine-friendly. You can feed it into pipelines, reuse it across projects, or combine it with automation scripts.
In short, JSON makes your creative work portable and scalable.
Core Structure
Think of core structure as your JSON skeleton. Without this, the rest won’t stand.
- Organize logically: Break your prompt into sections such as scene, characters, camera, style, environment.
- Syntax discipline: Always use brackets {}, commas ,, and quotation marks "" correctly.
- Separate objects: Each visual element (like a character or camera setting) should be its own JSON object.
- Validate: Run your JSON through a validator before sending it to NanoBanana. Small typos can derail the entire output.
Example
{
"scene": "Bustling Las Vegas casino floor",
"characters": [
{
"name": "Lead Gambler",
"description": "Sharp tuxedo, glass of whiskey",
"pose": "leaning against roulette table",
"emotion": "confident smirk"
}
],
"camera": {
"angle": "overhead",
"focus": "roulette wheel",
"style": "cinematic"
}
}
Style Definition
Style tells NanoBanana how to dress your scene. Without it, you risk flat or generic output.
- Primary style: photorealistic, cinematic, documentary, hyperrealistic.
- Rendering quality: detailed, high-resolution, professional.
- Surface textures: authentic material details, wear patterns, natural imperfections.
- Lighting: natural, studio, dramatic, soft ambient.
Example
"style": {
"primary": "cinematic",
"rendering": "hyperrealistic",
"textures": ["authentic materials", "natural wear patterns"],
"lighting": "soft ambient"
}
Technical Specifications
NanoBanana respects technical direction. These are the camera and physics instructions that add believability.
- Camera settings: depth of field, focal length, aperture.
- Resolution: at least “high definition minimum.”
- Rendering: anti-aliasing, noise control, color depth standards.
- Physics: gravity effects, structural believability.
Example
"technical": {
"camera": {
"depth_of_field": "shallow",
"focal_length": "85mm",
"aperture": "f/1.8"
},
"resolution": "4K",
"physics": "realistic gravity and balance"
}
Material Properties
Details make the difference between a flat render and an image you could mistake for a photo.
- Skin: pores, natural imperfections, ethnic diversity.
- Fabric: thread patterns, drape, wear indicators.
- Surfaces: scratches, patina, oxidation, natural irregularities.
- Transparency: refraction, surface interactions.
Example
"materials": {
"skin": "visible pores, natural imperfections",
"fabric": "realistic thread patterns, soft drape",
"surfaces": "subtle scratches, authentic patina",
"transparency": "accurate refraction on glass"
}
Environmental Factors
A scene without environment feels like a cutout. Add atmosphere, weather, and seasonal cues.
- Atmosphere: distance haze, humidity, smoke, mist.
- Time/season: lighting differences between winter dusk and summer noon.
- Particles: dust, moisture, floating pollen for atmospheric depth.
Example
"environment": {
"atmosphere": "light haze with neon reflections",
"season": "summer night",
"particles": "floating dust motes"
}
Composition Controls
Composition decides where the viewer’s eye goes.
- Perspective: natural human vision, wide shot, macro.
- Framing: rule of thirds, golden ratio.
- Placement: balanced, aesthetically pleasing.
Example
"composition": {
"perspective": "wide shot with human eye realism",
"framing": "rule of thirds",
"placement": "lead character centered, roulette wheel offset"
}
Quality Keywords
Keywords help NanoBanana fine-tune results.
- Include: hyperrealistic, photographic quality, natural lighting, authentic textures.
- Avoid: digital artifacts, unrealistic proportions, oversaturated colors.
- Reference: “National Geographic,” “professional photography standards.”
Example
"keywords": {
"include": ["hyperrealistic", "authentic textures"],
"avoid": ["oversaturated colors", "digital artifacts"],
"reference": "professional photography standards"
}
A Full Example
Here’s everything combined into a single NanoBanana JSON:
{
"scene": "High-end Las Vegas casino floor",
"characters": [
{
"name": "Lead Gambler",
"description": "Elegant tuxedo, whiskey glass in hand",
"pose": "resting at roulette table",
"emotion": "confident smirk"
}
],
"camera": {
"angle": "overhead cinematic shot",
"focus": "roulette wheel",
"style": "cinematic"
},
"style": {
"primary": "cinematic",
"rendering": "hyperrealistic",
"textures": ["authentic materials", "natural wear patterns"],
"lighting": "soft ambient"
},
"technical": {
"camera": {
"depth_of_field": "shallow",
"focal_length": "85mm",
"aperture": "f/1.8"
},
"resolution": "4K",
"physics": "realistic gravity and balance"
},
"materials": {
"skin": "visible pores, natural imperfections",
"fabric": "realistic thread patterns, soft drape",
"surfaces": "subtle scratches, authentic patina",
"transparency": "accurate refraction on glass"
},
"environment": {
"atmosphere": "light haze with neon reflections",
"season": "summer night",
"particles": "floating dust motes"
},
"composition": {
"perspective": "wide shot with human eye realism",
"framing": "rule of thirds",
"placement": "lead character centered, roulette wheel offset"
},
"keywords": {
"include": ["hyperrealistic", "authentic textures", "natural lighting"],
"avoid": ["digital artifacts", "oversaturated colors"],
"reference": "professional photography standards"
},
"duration": 8
}
Common Pitfalls to Avoid
- Overstuffing keywords – Too many descriptive words can confuse rather than help.
- Mixing styles – Don’t combine “cinematic” with “comic book” unless that clash is intentional.
- Ignoring structure – Missing commas or mismatched brackets will break the prompt.
- Forgetting environment – Scenes without atmosphere feel flat, even if characters are perfect.
Advanced Tips
- Layer styles carefully – You can use cinematic + hyperrealistic, but avoid contradictory mixes.
- Iterate in pieces – Test a character prompt first, then add environment, then technical specs.
- Use references – Phrases like “National Geographic style” or “studio photography standards” anchor quality.
- Think modular – Save your JSON blocks (style, environment, materials) and swap them for new projects.
Conclusion
NanoBanana isn’t about typing flowery descriptions. It’s about giving the AI a blueprint in JSON that is structured, logical, and precise. Once you master the categories—core structure, style, technicals, materials, environment, composition, and keywords—you can generate cinematic, photorealistic images that hold up under scrutiny.
Start simple, validate your syntax, and build complexity step by step. Soon you’ll have a library of reusable JSON templates that can spin out endless variations.
NanoBanana speaks JSON. If you learn its language, you control the output.
Comments
Post a Comment