Prompt Guide: Google Nano Banana
Google Nano Banana 2 and Nano Banana Pro are advanced image generation and editing models powered by Google's Gemini 3 architecture. They apply deep reasoning to fully comprehend prompts, support pro-level typography, translation, and custom aspect ratios.
This guide details the prompting frameworks and creative controls needed to get the best results from Google Nano Banana.
1. Core Prompting Best Practices
- Start with a Verb: Begin your prompt with a strong verb that dictates the primary operation (e.g.,
"Create a detailed portrait...","Transform this sketch...","Edit the background..."). - Use Positive Framing: Describe what should be in the scene rather than what to exclude. Instead of
"no cars, no crowds,"write"empty street, deserted pathway." - Direct like a Creative Director: Control the lighting, camera lens, color grading, and textures using concrete photographic and cinematic terms rather than generic quality words.
- Conversational Iteration: Refine and adjust your images step-by-step using natural follow-up prompts rather than overloading everything into a single initial prompt.
2. The Four Prompting Frameworks
Framework 1: Text-to-Image Generation (Without References)
When starting with a blank canvas, write a cohesive narrative description instead of a simple keyword list.
$$\text{Formula: } [\text{Subject}] + [\text{Action}] + [\text{Location/Context}] + [\text{Composition}] + [\text{Style}]$$
Example Prompt: A striking fashion model wearing a tailored brown dress, sleek boots, and holding a structured handbag. Posing with a confident, statuesque stance, slightly turned. A seamless, deep cherry red studio backdrop. Medium-full shot, center-framed. Fashion magazine style editorial, shot on medium-format analog film, pronounced grain, high saturation, cinematic lighting effect.
Framework 2: Multimodal Generation (With References)
Nano Banana allows you to combine up to 14 reference images (such as layouts, sketches, or textures) to guide the final visual output.
$$\text{Formula: } [\text{Reference Images}] + [\text{Relationship Instruction}] + [\text{New Scenario}]$$
Example Prompt: Using the attached napkin sketch as the structure and the attached fabric sample as the texture, transform this into a high-fidelity 3D armchair render. Place it in a sun-drenched, minimalist living room.
Framework 3: Image Editing & Inpainting
When editing an existing image, guide the model by specifying exactly what changes and what must stay identical.
- Conversational Editing: Describe the change directly (e.g.,
"Remove the man from the photo"). - Semantic Masking: Prompt specifically for the object or area to edit. Be explicit about keeping the surrounding composition identical (e.g.,
"Change the color of the car to red, keep the background and people exactly the same").
Framework 4: Text Rendering and Multilingual Localization
To render crisp, legible text inside your designs, or translate text overlays into other languages (supporting over 10 languages):
- Wrap in Quotes: Always put the exact letters you want to render in double quotes (e.g.,
"GLOW"or"New York"). - Describe the Typography: Name the font style or describe it (e.g.,
"Century Gothic font","Brush Script font","bold white Impact font"). - Use the Text-First Hack: Converse with the AI assistant first to generate the text content, and then instruct it to render the final image with that specific text.
3. Creative Studio Controls
A. Lighting Design
- Studio Setups: Ask for a
"three-point softbox setup"for clean product shots. - Dramatic Effects: Prompt for
"Chiaroscuro lighting with harsh, high contrast"or"Golden hour backlighting creating long shadows."
B. Camera, Lens, and Focus
- Hardware DNA: Dictate specific cameras to change the look (e.g.,
"GoPro"for action distortion,"Fujifilm"for warm organic colors,"disposable camera"for raw nostalgic flash). - Lenses & Aperture: Specify
"shallow depth of field (f/1.8)"for blurred backgrounds, a"wide-angle lens"for landscape scale, or a"macro lens"for extreme close-up details.
C. Color Grading and Film Stock
- Vintage Looks: Request
"as if on 1980s color film, slightly grainy"or"classic Kodachrome tones." - Modern Grades: Prompt for
"Cinematic color grading with muted teal tones"or"vibrant high-saturation palettes."
D. Materiality and Texture
- Define the exact physical surfaces of objects (e.g.,
"navy blue tweed jacket," "ornate elven plate armor etched with silver leaf," "minimalist matte ceramic coffee mug").