Nano Banana 2: The Complete Prompting Guide

For Creative Professionals, Enthusiasts, and Everyone in Between

Welcome to Nano Banana 2

Picture this: you need a website hero banner, six product shots, three social media posts, and a quick draft for a campaign concept — all before your afternoon meeting. You don't want to wait. You don't want to drain your budget. And you definitely don't want to compromise on quality.

That's exactly why Nano Banana 2 (NB2) exists.

Launched on February 26, 2026 and built on Google DeepMind's Gemini 3.1 Flash Image architecture, NB2 is the workhorse of the Nano Banana image generation family. It's not the most powerful model in the lineup — that's Nano Banana Pro's job. But NB2 delivers fast, beautiful, highly capable image generation at a fraction of the cost, and it comes with two tricks that no other model has: ultra-extreme aspect ratios (perfect for web banners and tall posters) and live Image Search (so it can pull real-world visual references on the fly).

For creative professionals, NB2 is the model you'll have open all day. It's where you sketch, iterate, test compositions, crank out product series, and explore creative directions before committing to a Pro-quality final. Think of it as your studio assistant — fast, reliable, and surprisingly talented.

Who this guide is for: Designers, photographers, marketers, content creators, social media managers, e-commerce professionals, illustrators, and anyone who creates images regularly and wants to work smarter. You don't need to be an engineer. You just need to know what to ask for — and this guide will show you exactly how.

Quick Start: 3 Copy-Paste Prompts to Try Right Now

If you want to jump in before reading anything else, here are three production-ready prompts that showcase what NB2 does best. Copy them directly, run them, and tweak from there.

1. Instant E-Commerce Product Shot

A clean product photograph of a white ceramic candle jar with a wooden lid, sitting on a light marble surface. Soft studio lighting from the upper left, subtle shadow underneath. Minimalist lifestyle aesthetic, slightly warm color tone. 1:1 aspect ratio, 2K resolution. Avoid: busy backgrounds, harsh shadows, extra objects.

2. Sweeping Web Banner (NB2-exclusive ultra-wide ratio)

An 8:1 panoramic web banner showing a misty mountain range at sunrise. Layers of blue-gray peaks receding into distance, a warm amber sky on the far left, dense pine forest in the foreground. Cinematic, serene, suitable for a travel brand homepage header. No text. 2K resolution.

3. Real-World Location (powered by NB2's Image Search)

The Grand Canal in Venice, Italy at blue hour, just after sunset. Gondolas moored along the canal, the Rialto Bridge lit with warm lights in the middle distance. Reflections shimmer on the water. Shot on a Sony A7R V with 35mm lens, f/2.8. Photorealistic travel photography style. 16:9 aspect ratio, 4K resolution.

Table of Contents

1. Model Identity & What Makes NB2 Special
2. Technical Specifications
3. NB2 vs. Nano Banana Pro: Which Should You Use?
4. NB2-Exclusive Features (Deep Dive)
5. The Golden Rules of Prompting
6. Prompt Structure and Formulas
7. Prompt Anatomy: Every Component Explained
8. Text Rendering
9. Image Editing Workflows
10. Multi-Image and Character Consistency Workflows
11. Resolution and Format Selection
12. Creative Director Controls
13. Production-Ready Example Prompts
14. Common Mistakes → Better Versions
15. Prompt Building Workshop
16. NB2 vs. Pro Decision Flowchart
17. Cheat Sheet

1. Model Identity

NB2 is built on the Gemini 3.1 Flash Image model from Google DeepMind — the same underlying family as Nano Banana Pro, but optimized for speed, cost, and accessibility rather than maximum aesthetic ceiling.

What does "free tier included" mean for you? It means you can start using NB2 today without a paid subscription. The free tier gives you access to the same model — just with usage limits. For heavy professional use, the API is the cost-effective path at roughly $0.039 per image.

2. Technical Specifications

These are the hard numbers. Keep them handy when planning projects — particularly the aspect ratios, which are one of NB2's biggest differentiators.

A Note on the Context Window

131,072 input tokens is a large context window — roughly 100,000 words of text. In practice, this means you can feed NB2 enormously detailed briefs, multiple reference images, style guides, brand guidelines, and long character descriptions all at once without hitting limits. This is significantly larger than Nano Banana Pro's 65,536-token window, making NB2 paradoxically the better choice for complex multi-asset briefs even though Pro handles individual images more artfully.

3. NB2 vs. Nano Banana Pro

This is the question creative professionals ask most often. Here's the full side-by-side comparison — and the short answer: use NB2 for most work, and Pro only when you need its specific advantages.

The practical reality: NB2 handles 80–90% of professional creative work with excellent results. The gap between NB2 and Pro is real, but it's a gap you'll notice most in highly stylized hero shots for premium brand campaigns — not in social posts, product photography, banners, or illustration work where NB2 excels.

The detailed decision guide is in Section 16.

4. NB2-Exclusive Features (Deep Dive)

These two features aren't available anywhere else in the Nano Banana family. Understanding them is key to getting unique value from NB2.

4.1 Ultra-Wide and Ultra-Tall Aspect Ratios

NB2 supports 14 aspect ratios, including 4 extreme formats that no other model offers:

Why does this matter? Standard image generators only produce images in roughly rectangular or square proportions. If you need a web banner that spans the full width of a desktop screen, or a vertical scrolling infographic for Instagram Stories, you'd normally have to generate a wider image and crop — losing content, resolution, and composition control. NB2 is designed from the ground up to think in these extreme proportions, composing the scene to fill the entire canvas correctly.

When you need an extreme aspect ratio, NB2 is the only option. Always specify the ratio directly in your prompt:

• "An 8:1 panoramic banner showing..."
• "A 1:8 vertical story image..."
• "A 4:1 website header featuring..."

4.2 Image Search Tool

This is NB2's most underappreciated feature. When your prompt references real-world things — landmarks, specific products, brand identities, architectural styles, known locations — NB2 doesn't just rely on what it learned during training. It actively searches Google's image database to pull current, accurate visual references.

What this means in practice:

How to leverage it in your prompts:

• Name landmarks specifically: "the Sydney Opera House", "the Guggenheim Bilbao", "the Flatiron Building"
• Name products precisely: "a 2025 BMW M3 Competition", "the iPhone 16 Pro", "Nike Air Force 1"
• Name brands by full name: "the Google logo", "the Starbucks siren", "the Apple wordmark"
• Reference architectural styles with examples: "Gaudí-style organic architecture like Casa Batlló"

Image Search is especially powerful for: travel content, product mockups, brand placements, architectural visualization, and any work where visual accuracy to a real-world reference matters.

5. The Golden Rules

These principles apply to all Nano Banana models, but they're worth understanding deeply before diving into NB2-specific techniques.

Rule 1: Write Like You're Briefing a Creative Director

Use complete sentences, not keyword lists. NB2 understands natural language and responds better to well-formed descriptions than to comma-separated tags.

❌ portrait woman sunlight bokeh moody cinematic ✅ A portrait of a woman in her mid-30s, caught in afternoon sunlight streaming through a window, soft bokeh in the background, moody and cinematic mood.

Rule 2: Describe What You Want, Not What to Avoid

Frame your prompt positively. Instead of saying what you don't want, describe what you do want. (The exception: a short exclusion list at the end works well for photorealism quality control.)

❌ A product shot — no messy background, not overexposed, avoid shadows ✅ A clean product shot on a seamless white background, perfectly exposed, with only a soft natural shadow beneath the product.

Rule 3: Edit Conversationally

You don't need to get it perfect in one shot. NB2 is excellent at iterative refinement. Generate something 80% correct, then use follow-up prompts like:

• "Keep everything the same but change the lighting to blue neon"
• "Make the sky more dramatic with storm clouds"
• "Shift the color palette to warm earth tones"

Rule 4: Be Specific Across Four Dimensions

The most common reason prompts produce generic results is under-specification. Cover all four:

1. Subject — Who or what, with specific details
2. Environment — Where, with specific atmosphere
3. Light — Quality, direction, color temperature
4. Style — Photography, illustration, rendering, art movement

Rule 5: Control the Camera

Use photographic and cinematic language. NB2 understands and responds to lens lengths, apertures, shot sizes, and angles. These controls give you precise compositional guidance.

Rule 6: Start with a Strong Verb

Your prompt's first word signals the primary operation. This helps NB2 orient to your intent immediately:

• "Generate...", "Create...", "Photograph..." for new images
• "Edit...", "Modify...", "Transform..." for changes
• "Blend...", "Combine...", "Merge..." for multi-image work

6. Prompt Structure and Formulas

Formula 1: Text-to-Image (Most Common)

[Subject] + [Action/State] + [Location/Context] + [Composition] + [Style]

Example breakdown:

• Subject: "A 1920s jazz musician"
• Action/State: "playing trumpet, eyes closed, completely lost in the music"
• Location/Context: "in a smoky underground club in Harlem"
• Composition: "medium shot, slightly low angle, tight framing"
• Style: "black and white documentary photography, high contrast, film grain"

Why This Works: This formula mirrors how a film director briefs a cinematographer — subject first (who), then action (what's happening), then environment (where), then technical controls (how to shoot it). NB2 processes context in layers, and feeding it this structure prevents ambiguity at each level before moving to the next.

Formula 2: Multimodal with Reference Images

[Reference images] + [Relationship instruction] + [New scenario]

Example:

[Image 1: reference of a character]
[Image 2: reference of a style]

Using Image 1 as the character and Image 2 as the artistic style, create a scene where the character is walking through a neon-lit Tokyo street at night. Keep the character's facial features exactly as shown in Image 1. Apply the brushstroke texture and color saturation from Image 2 throughout.

Why This Works: NB2 needs to understand what role each reference plays. Without explicit role assignments, it may blend all references together or pick one arbitrarily. The "relationship instruction" is the linchpin — it gives each visual reference a job to do.

Formula 3: Web Search Generation (NB2 Exclusive)

[Source/Search request] + [Analytical task] + [Visual translation]

Example:

Search for current images of the interior of the Louvre Museum's Denon Wing. Using those references, create a wide-angle architectural photograph showing the gallery at twilight, with soft spotlights on the artworks and long shadows across the marble floors. No visitors present. 21:9 cinematic aspect ratio.

Why This Works: This formula explicitly activates NB2's Image Search tool. By framing the first part as a search request, you signal that you need real-world accuracy. The analytical task (what to focus on) and visual translation (how to render it) then give NB2 clear direction on what to do with those references. This combination is only possible in NB2.

7. Prompt Anatomy: Every Component Explained

Think of a great prompt as having seven organs, each with a specific function. You don't need all seven in every prompt — but the more you include, the more control you have.

Building the Full Prompt:

A weathered fisherman in his 60s, weathered skin and salt-and-pepper stubble, mending a fishing net on a wooden dock. A foggy New England harbor at dawn surrounds him, with fishing boats barely visible through the mist. Medium shot, slightly low angle, 50mm lens at f/2.8 with shallow depth of field. Soft diffused light filters through the morning fog, creating a melancholic, quiet mood. Documentary photography style in the tradition of Steve McCurry, National Geographic aesthetic. Avoid: bright colors, modern objects, digital artifacts.

That's a complete, production-ready prompt built from the seven components. Notice how each piece adds a layer of specificity without being redundant.

8. Text Rendering

NB2 shares Nano Banana Pro's excellent text rendering capability. Getting clean, accurate text in your images requires following a few conventions.

The Four Rules of Text Rendering

1. Always put text in quotation marks

A minimalist poster with the words "BE HERE NOW" in bold Helvetica, centered.

The quotation marks tell NB2 exactly what string to render. Without them, it may paraphrase or stylize the text.

2. Describe the font/style explicitly

• "bold sans-serif" — clean modern look
• "hand-painted sign lettering" — artisanal, imperfect
• "neon cursive" — glowing, scripted
• "vintage slab serif, slightly distressed" — retro/authentic
• "engraved metal lettering" — premium, dimensional

3. Multilingual text works NB2 accurately renders text in 10+ languages. Simply include the text in the target language within quotes and specify the language if context is unclear: "the word 'BIENVENIDO' in hand-carved wooden letters"

4. The Text-First Hack for Complex Copy For longer text that must be absolutely accurate (a tagline, a legal disclaimer, product copy), use a two-step approach:

1. Generate and confirm the exact text in a conversational message first
2. Then say: "Now incorporate this text into an image: [paste confirmed copy here]"

This works because NB2 maintains context within a session, so the "confirmed" text becomes an anchor it won't deviate from.

9. Image Editing Workflows

NB2's editing capabilities are conversational and semantic — meaning you describe changes in plain English, and it applies them intelligently without you needing to draw masks or use selection tools.

Semantic Masking (Inpainting)

You can tell NB2 to change a specific part of an image using natural language descriptions of what to change:

• "Remove the person on the left and extend the sidewalk naturally"
• "Keep the composition but change the lighting to golden hour"
• "Replace the background with a neon-lit city street at night"
• "Add a steaming coffee mug to the table in the lower right corner"
• "Change the model's jacket from red to olive green"

Style Transfer

Apply the visual character of one image to another:

• "Recreate this photo as a watercolor painting with loose brushstrokes"
• "Apply the color palette and moody lighting from Image B to Image A"
• "Transform this product shot into a vintage 1970s advertisement style"

Iterative Refinement

NB2 retains context within a session, so each edit builds on the last. A typical professional workflow might look like:

1. "Create a rainy city street at night, cinematic, 16:9" → Generate base
2. "Make the reflections on the pavement more vivid" → Enhance
3. "Add a neon sign reading 'OPEN' in the upper right background" → Add element
4. "Darken the shadows in the foreground" → Mood adjustment
5. "Change to 4K resolution for final output" → Upscale

Each step is a short, natural instruction. This workflow is often faster and produces better results than trying to specify everything in one giant prompt upfront.

10. Multi-Image and Character Consistency Workflows

Character Consistency (Up to 5 Characters)

NB2 can maintain consistent facial features and appearance for up to 5 characters across a series of images — ideal for storyboards, comics, branded illustrations, and sequential content.

Best practices for character consistency:

1. Use clear reference images — Front-facing, well-lit, with the character's distinctive features clearly visible
2. Assign explicit names — Don't say "the woman in Image 1." Name them: "Maya," "Character A," "the detective"
3. State consistency explicitly — "Keep Maya's facial features, hairstyle, and skin tone exactly consistent with her reference image throughout all scenes"
4. Describe what can vary — Tell NB2 what IS allowed to change: "Her clothing and expression can vary per scene, but her face must remain identical"

Example character consistency prompt:

Using the attached reference images for Character A (Maya, the engineer) and Character B (Leon, the pilot):

Scene 1: Medium shot of Maya reviewing a holographic display in a spacecraft cockpit. Focused expression. Blue ambient lighting.
Scene 2: Wide shot of Leon at the helm of the spacecraft, stars visible through the front window. Dramatic backlighting.
Scene 3: Two-shot of Maya and Leon looking at each other tensely. Red emergency lighting.

Keep both characters' facial features, hairstyles, and skin tones exactly consistent with their reference images across all three scenes. Cinematic 2.39:1 aspect ratio, 2K resolution each.

Image Fusion (Up to 14 Object References)

NB2 can blend up to 14 reference images into a single coherent composition. The key is assigning each reference a clear role:

Using the provided reference images:
- Image A provides the character (keep facial features exact)
- Image B provides the art style (brushstroke texture and color palette)
- Image C provides the background environment
- Image D provides the clothing style

Create a scene of the character walking through a market, rendered in the art style, wearing the clothing style, with the background adapted to a daytime version of Image C.

11. Resolution and Format Selection

Choosing the right resolution is as much a workflow decision as an aesthetic one. Higher resolution costs more and takes longer — and for drafts, that's wasted time and money.

The Smart Iteration Workflow

Draft phase: 0.5K — Generate 5–10 variations quickly and cheaply Refinement phase: 1K — Narrow down to 2–3 strong concepts and refine Production phase: 2K — Create polished versions for digital use Final output: 4K — Only for the one(s) going to print or high-res display

Specifying Format in Prompts

You can request aspect ratios directly in natural language:

• "A 9:16 vertical social story"
• "An 8:1 panoramic web banner"
• "A cinematic 21:9 wide shot"
• "A 1:1 square product image"
• "A 4:1 website header"

12. Creative Director Controls

These are the vocabulary terms that give you precise control over the visual output. Learn these phrases and you'll have a much larger range of expressive control.

Lighting

Camera and Lens

Film Stock and Color Grading

13. Production-Ready Example Prompts

These prompts are designed to work immediately — copy, run, and adjust to your needs. Each includes a brief note on the key choices made.

Ultra-Wide Banners (8:1 and 4:1)

Solar System Banner (original from guide)

An 8:1 panoramic banner showing the complete solar system in order from the Sun to Neptune. Each planet is rendered at accurate relative scale with labeled names in clean white Helvetica font. Deep space background with subtle nebula colors. Scientific illustration style with modern data visualization aesthetics. 2K resolution.

Note: The 8:1 ratio is perfect for this use case — the planets naturally spread across a wide linear composition. The Helvetica label instruction keeps text clean and authoritative.

Mountain Sunrise Travel Banner

An 8:1 panoramic web banner showing a dramatic mountain range at sunrise, photographed from a high alpine meadow. Layers of peaks fade into haze in the distance, with warm golden light catching snow-covered summits on the right. Foreground wildflowers in soft focus. Suitable for a luxury travel brand homepage. No text. 2K resolution, cinematic color grade.

Note: "Suitable for a luxury travel brand" is a style shorthand that communicates overall aesthetic more efficiently than exhaustively listing every quality parameter.

Brand Timeline Header

A 4:1 horizontal website header showing a clean timeline graphic from 1985 to 2025. Each decade marked with a subtle vertical line and year label in thin sans-serif font. Gradient background from deep navy on the left to bright cobalt on the right. Minimalist, corporate design aesthetic. No icons or illustrations — typography and geometry only. 2K resolution.

Note: Using 4:1 for a timeline is a natural fit since timelines are inherently horizontal. The explicit "no icons" exclusion prevents NB2 from decorating beyond your intended design.

Cinematic City Skyline Banner

An 8:1 panoramic banner showing the New York City skyline at blue hour, photographed from across the Hudson River in New Jersey. The Empire State Building and One World Trade Center visible, warm office lights glowing in hundreds of windows, deep blue sky with the last trace of sunset on the horizon. Photorealistic, shot with a 200mm telephoto lens for compressed perspective. 4K resolution.

Note: Naming the real location activates NB2's Image Search for accurate skyline detail. The 200mm telephoto instruction compresses the skyline for a graphic, poster-like quality.

Abstract Tech Banner

A 4:1 website header for a software company. Abstract network visualization: glowing nodes connected by thin luminous lines against a deep black background. Cool blue and purple color scheme, subtle particle effects. Clean, technical, futuristic aesthetic. The text "CONNECT EVERYTHING" in bold uppercase white Grotesk font appears centered. 2K resolution.

Note: Combining a visual description with placeholder headline copy allows the design and text to be composed together rather than added separately.

E-Commerce Product Photography

Ceramic Pour-Over (original from guide)

A clean, minimalist product shot of a matte ceramic pour-over coffee dripper in sage green. White seamless background, soft studio lighting from the upper left with a subtle fill light. Include a small shadow beneath. The text "BREW" appears in thin, uppercase Futura font below the product. 1:1 aspect ratio, 2K resolution. Avoid: any other objects, colored backgrounds, harsh shadows.

Luxury Skincare Product

A product photograph of a small amber glass serum bottle with a gold dropper cap, standing upright on a white marble surface. Soft, even studio lighting with a subtle reflection in the marble. A single dried botanical sprig is placed to the left for scale and context. Clean, luxury skincare aesthetic. The label text "RENEW" is visible in embossed gold serif lettering. 1:1 aspect ratio, 2K resolution. Avoid: busy backgrounds, props that compete with the product, harsh shadows.

Note: The single botanical prop adds context and elegance without cluttering — specify it explicitly so NB2 keeps restraint.

Athletic Sneaker

A studio product photograph of a white leather low-top sneaker with a thick gum sole, shot at a 45-degree angle from slightly above. Clean white background, dramatic key light from the right creating a defined shadow to the left. Shot on a Hasselblad medium format camera, sharp throughout with no depth of field blur. The shoe is positioned on a subtle light grey platform. 3:2 aspect ratio, 4K resolution.

Note: The 45-degree-from-above angle is the classic sneaker photography angle — specifying it directly ensures the right composition instead of a flat side-on view.

Food Product Flatlay

A flatlay product photograph of a kraft paper bag of artisan coffee beans, surrounded by scattered whole coffee beans on a dark wood surface. Overhead shot, exactly 90 degrees, slight warm natural light from the upper right creating texture in the wood grain. The coffee bag shows a hand-drawn illustrated label with the text "SINGLE ORIGIN" visible. Warm, artisan, specialty coffee brand aesthetic. 1:1 aspect ratio, 2K resolution.

Note: Flatlay requires the exact-90-degree-overhead specification; without it, NB2 may default to a 3/4 angle.

Watch Product Lifestyle Shot

A lifestyle product photograph of a classic stainless steel watch with a navy canvas strap, worn on a man's wrist. He is sitting at a weathered wood table with a half-drunk espresso cup and open leather notebook in the background, slightly out of focus. Natural morning light from a large window to the left. Warm, aspirational, Muji-inspired aesthetic. 3:2 aspect ratio, 2K resolution, shot on Leica Q3, 28mm f/1.7.

Note: The out-of-focus background props are specified deliberately — they create lifestyle context without distracting from the watch. The Leica camera reference sets a particular color rendering aesthetic.

Social Media

Instagram Reel Cover (9:16)

A 9:16 vertical social media image for an interior design brand. A bright, airy living room in the Japandi aesthetic: natural wood tones, white walls, minimal furniture, a single large houseplant in a terracotta pot by a window. Afternoon sunlight streaming in, casting long rectangular shadows. Clean and aspirational. 1K resolution.

Note: Naming the style ("Japandi") is a powerful shorthand that communicates an entire aesthetic vocabulary. NB2 understands it.

Instagram Square Post (1:1)

A 1:1 square Instagram post graphic. Bold abstract expressionist composition in an electric blue, burnt orange, and cream color palette. Large brushstrokes dominate the frame. In the lower third, the text "SLOW DOWN" in thin white sans-serif type, left-aligned, with plenty of breathing room. Fine art meets graphic design aesthetic. 2K resolution.

Note: Placing the text "in the lower third" is specific enough that it won't float randomly — and the "breathing room" instruction prevents overcrowding.

Pinterest Tall Pin (2:3)

A 2:3 tall Pinterest-style recipe pin. A top-down flat lay of homemade sourdough bread ingredients: flour in a wooden bowl, active starter in a glass jar, salt, and a linen cloth on a dark blue stone surface. Warm moody natural lighting. At the top, a clean white text box with the title "SOURDOUGH FROM SCRATCH" in a bold serif headline font. Professional food photography aesthetic. 2K resolution.

LinkedIn Thought Leadership Post

A horizontal 16:9 image for a LinkedIn post about the future of AI. Abstract visualization: a human hand reaching toward a glowing network of interconnected nodes, rendered as a high-quality 3D illustration against a deep space background. Blues, purples, and warm white light. Professional, aspirational, slightly futuristic without feeling dystopian. No text. 1K resolution.

Note: "Without feeling dystopian" is a mood qualifier that effectively excludes dark, threatening visual associations — a real problem with AI imagery if you don't specify.

Twitter/X Card (16:9)

A 16:9 illustration for a tweet about climate data. Clean data visualization aesthetic: a stylized rising temperature graph as the dominant visual element, rendered in a warm red-to-amber gradient. Dark navy background. The numbers "1.5°C" appear large and bold in white in the upper right. Modern infographic illustration style, suitable for editorial use. 1K resolution.

Real-World Locations (Image Search)

Fushimi Inari Shrine (original from guide)

The Fushimi Inari Shrine in Kyoto, Japan at golden hour. A long perspective shot looking through the iconic orange torii gate tunnel. Warm light filters through gaps between gates. A single person in traditional kimono walks ahead. Photorealistic, shot on Fujifilm X-T5 with 23mm lens. 16:9 aspect ratio, 4K resolution.

Grand Canal, Venice

The Grand Canal in Venice, Italy at blue hour, just after sunset. Gondolas moored along the canal, the Rialto Bridge lit with warm lights in the middle distance. Reflections shimmer on the still water. Shot on a Sony A7R V with 35mm lens, f/2.8. Photorealistic travel photography style. 16:9 aspect ratio, 4K resolution.

Note: "Blue hour" is more precise than "night" or "evening" — it specifies a specific 20-minute window of post-sunset light that photographers prize for its deep blue tones and soft glow.

Santorini Sunset

The village of Oia in Santorini, Greece at sunset. The iconic white-washed cubic buildings with blue-domed churches cascade down the volcanic caldera cliff. The setting sun paints the sky in shades of orange, peach, and magenta. Shot from the famous sunset viewpoint, 35mm equivalent, photorealistic travel photography. 21:9 cinematic aspect ratio, 4K resolution.

Note: Specifying "the famous sunset viewpoint" helps NB2 understand the specific angle, since Oia is most recognizable from that particular vantage point.

Shibuya Crossing, Tokyo

The Shibuya Crossing in Tokyo at night. The scramble crossing packed with people moving in all directions, motion blur applied to pedestrians with a 1-second exposure effect. Giant neon advertising screens reflected on the wet asphalt. Shot from an elevated perspective, slightly above eye level, looking straight down the main crossing. Photorealistic, urban documentary photography. 16:9 aspect ratio, 4K resolution.

Note: The "1-second exposure effect" tells NB2 to render the motion blur you'd get with a slow shutter — a powerful technique for capturing the energy of crowded places.

Amalfi Coast Road

The famous Amalfi Coast road on the Italian Riviera, photographed from a drone at 200 meters altitude. The narrow road winds along sheer cliffs above turquoise Mediterranean water, with colorful stacked villages visible below. Bright summer midday light, deep blue sky with a few white clouds. Photorealistic aerial photography. 21:9 cinematic aspect ratio, 4K resolution.

Character Consistency

Multi-Scene Storyboard (original from guide)

Using the attached reference images for Character A (the detective) and Character B (the informant):

Image 1: Medium shot of Character A sitting in a dimly lit diner booth, looking suspicious. Film noir style.
Image 2: Close-up of Character B sliding an envelope across the table. Dramatic side lighting.
Image 3: Wide shot of both characters from across the diner. Neon sign reading "OPEN" visible through the window.
Image 4: Character A alone, opening the envelope under a street lamp outside. Rain falling.

Keep both characters' facial features exactly consistent with their reference images. Cinematic 2.39:1 aspect ratio, 2K resolution.

Children's Book Sequence

Using the attached reference image for Elara (a 7-year-old girl with bright red hair, freckles, and green eyes, wearing a yellow raincoat):

Illustration 1: Elara discovers a tiny glowing door at the base of an old oak tree. Wide-eyed with wonder. Lush forest, dappled light.
Illustration 2: Elara crawls through the tiny door, which has expanded to fit her. Magical purple light beyond.
Illustration 3: Elara stands in a meadow filled with oversized luminous mushrooms and friendly foxes. Delighted expression.
Illustration 4: Elara asleep at the base of the oak tree, a small golden flower in her hand. Was it a dream?

Keep Elara's appearance exactly consistent across all illustrations. Children's picture book illustration style, warm and magical color palette. 4:3 aspect ratio, 2K resolution.

Note: Detailed physical description of the character ("bright red hair, freckles, green eyes, yellow raincoat") in the reference instruction ensures consistency even if the reference image is approximate.

Brand Mascot Series

Using the attached reference image of the brand mascot "Zip" (a geometric robotic fox with orange and white coloring, expressive blue LED eyes, and a sleek modern design):

Pose 1: Zip waving enthusiastically at the viewer, front-facing. White background.
Pose 2: Zip sitting and looking thoughtful, one paw under chin. White background.
Pose 3: Zip running with a tablet under one arm, determined expression. White background.
Pose 4: Zip sleeping curled up with a "ZZZ" speech bubble. White background.

Keep all of Zip's design details exactly consistent. Flat vector illustration style, clean lines, suitable for brand asset use. 1:1 aspect ratio, 2K resolution. Avoid: photorealistic rendering, shadows, background elements.

Note: For brand mascots, specifying "suitable for brand asset use" communicates that you need clean edges and consistent rendering appropriate for production assets.

Illustrations and Infographics

Scientific Infographic Strip (8:1)

An 8:1 horizontal infographic strip illustrating the human digestive system from mouth to colon. Each organ labeled with clean sans-serif text. Smooth medical illustration style, slightly stylized (not photorealistic), pastel blue and coral color palette on a white background. Suitable for a health education website. Clear visual flow from left to right. 2K resolution.

Vertical Infographic (1:4)

A 1:4 vertical infographic about the water cycle. From top to bottom: cloud formation with rain, water flowing into rivers, groundwater seeping into soil, evaporation rising from lakes back to clouds. Each stage labeled and connected by elegant arrows. Nature illustration style, soft watercolor textures, earthy blue and green palette. Suitable for a school educational poster. 2K resolution.

Note: The 1:4 ratio is perfect for vertical infographics that flow from top to bottom — a natural fit for process diagrams.

Character Illustration: Fantasy Warrior

A full-body character illustration of a female warrior in ornate silver and jade armor. She stands in a confident pose, sword resting on her shoulder, helm under one arm. Detailed fantasy illustration style reminiscent of cover art for epic fantasy novels. Rich, saturated colors with detailed highlights and shadows. The background is a misty mountain pass, kept purposefully blurred to emphasize the character. 2:3 portrait aspect ratio, 4K resolution.

Data Visualization Banner

A 4:1 wide infographic banner showing global renewable energy adoption from 2000 to 2025. A line graph with multiple colored lines for solar, wind, hydro, and nuclear. Clean modern data visualization design, white background, a restrained color palette (teal, amber, coral, slate blue), thin gridlines, year labels on the X-axis. The title "THE RENEWABLE REVOLUTION" in bold dark sans-serif at the left. 2K resolution.

Editorial Illustration

An editorial illustration for a magazine article about remote work and loneliness. A person sits at a glowing laptop in a small, cozy apartment at night. Outside the window, a city glows with lights and implied human activity. The person is surrounded by space — the room is too big, too quiet. The contrast between the warm screen and the cold city outside should be the emotional core. Semi-flat illustration style, muted palette of navy blue, warm amber, and pale grey. 3:2 aspect ratio, 2K resolution.

Note: Providing the emotional interpretation ("the contrast...should be the emotional core") guides NB2 toward the right mood rather than leaving visual metaphor to chance.

14. Common Mistakes → Better Versions

These before-and-after comparisons show the most common prompting errors and how to fix them. Each pair reveals a specific principle.

Mistake 1: Keyword soup

❌ Before:

woman portrait bokeh sunlight hair golden studio professional photography

✅ After:

A professional portrait of a woman in her early 30s, photographed in a naturally lit studio with warm afternoon sunlight streaming through a large window behind her. Shallow depth of field creates soft golden bokeh in the background. Shot on an 85mm f/1.4 lens. Warm, professional, editorial aesthetic.

Why it's better: Full sentences give NB2 the relationships between elements. In the keyword list, "sunlight" and "studio" conflict — NB2 doesn't know how to combine them. The after version resolves the conflict with "naturally lit studio with sunlight through a window."

Mistake 2: Negative framing

❌ Before:

A product photo of a watch. No messy background. Not overexposed. Don't include other products. No harsh lighting.

✅ After:

A clean product photograph of a watch on a seamless white background. Soft, even studio lighting with no harsh shadows. The watch is the only object in the frame, shot at a 3/4 angle from slightly above.

Why it's better: Every "don't" in the original describes an unwanted outcome without telling NB2 what the wanted outcome looks like. The after version replaces each negative with a positive description.

Mistake 3: Using 4K for everything

❌ Before:

[Generates 15 product photo variations at 4K for initial concept testing]

✅ After:

[Generates 15 variations at 0.5K to find the right composition, then regenerates the winner at 4K]

Why it's better: 0.5K generation costs roughly 1/4 of 4K and takes a fraction of the time. For exploration, the resolution difference doesn't matter — you're evaluating composition and lighting, not pixel quality. This workflow saves approximately $1.95 for every 15-variation exploration round.

Mistake 4: Forgetting to use Image Search for real places

❌ Before:

A famous Japanese temple with orange gates.

✅ After:

The Fushimi Inari Shrine in Kyoto, Japan — the famous Senbon Torii path of over ten thousand orange torii gates winding up Mount Inari. Photographed from inside the gate tunnel looking uphill, late afternoon light. Photorealistic. 16:9 aspect ratio.

Why it's better: The vague description gives NB2 no visual anchor and produces a generic result. The specific name triggers Image Search for accurate visual references. The additional details (looking uphill, inside the tunnel) specify the composition.

Mistake 5: Not specifying the aspect ratio for banners

❌ Before:

A wide web banner for our website showing a mountain landscape.

✅ After:

An 8:1 panoramic web banner showing a dramatic mountain landscape at dawn. The Dolomites in northern Italy, photographed from a valley looking up at jagged limestone peaks catching the first morning light. Horizontal composition optimized for web banner proportions. No text. 2K resolution.

Why it's better: "Wide" is subjective and will produce a 16:9 image at best. 8:1 is the instruction that unlocks NB2's exclusive ultra-wide format and produces a banner-ready composition.

Mistake 6: Generic style description

❌ Before:

A portrait photo. Make it look professional and cinematic.

✅ After:

A cinematic portrait of a man in his 50s with weathered features, silver-streaked hair, and deep-set eyes. Shot on a 70mm anamorphic lens, slight lens flare from practical lighting, teal and amber color grade. Dark background, single strong key light from the left. The mood of a quiet scene between action sequences in a prestige TV drama.

Why it's better: "Professional and cinematic" means different things to different people — and to NB2. Specific lens, color grade, lighting setup, and cultural reference (prestige TV drama) paint an unambiguous picture.

Mistake 7: Not using quotation marks for text

❌ Before:

A poster showing the word SALE in big red letters.

✅ After:

A retail promotion poster featuring the word "SALE" in oversized bold red Impact-style uppercase letters, centered on a white background. Clean, graphic, high-contrast. 2:3 portrait poster aspect ratio, 2K resolution.

Why it's better: Without quotes, NB2 may paraphrase or stylize the text ("SAVINGS!", "HUGE DEALS"). With quotes, it renders the exact string.

Mistake 8: Forgetting exclusions for photorealism

❌ Before:

A photorealistic portrait of a person on a busy street.

✅ After:

A photorealistic street portrait of a person in their mid-40s on a busy Tokyo shopping street. Natural ambient light, 35mm f/2 lens, documentary photography aesthetic. Avoid: extra fingers, deformed hands, distorted facial features, blurry eyes, artificial-looking skin texture, visible artifacts.

Why it's better: The exclusion list is specifically targeted at the common failure modes in AI portrait generation. Adding it significantly improves photorealism quality, especially for hands and faces.

15. Prompt Building Workshop

Here's how a professional prompt evolves from vague to refined. This is the real process — and following these five stages helps you understand why each element matters.

Brief: An image for a premium headphone brand's digital campaign. Feature their product in a lifestyle context.

Iteration 1: The Rough Idea

Headphones on a person listening to music.

What NB2 produces: A generic stock-photo-feeling image of someone wearing headphones, probably on a white background or in a neon-light bedroom. Technically correct, not useful.

Problem: No brand premium signal, no emotional context, no scene, no style.

Iteration 2: Add Context and Mood

A person wearing premium over-ear headphones, sitting in a comfortable space, listening to music with eyes closed. Warm, premium aesthetic.

What NB2 produces: Better — there's a mood. But "comfortable space" is vague, "warm, premium" doesn't mean much without specifics, and there's still no character or story.

Problem: The setting is still undefined, the character has no specificity, and "premium aesthetic" is too generic.

Iteration 3: Ground It in a Specific Scene

A young woman in her late 20s with short natural hair sits in a beautifully designed living room, wearing large black over-ear headphones. Eyes closed, a slight smile on her face. Natural light from tall windows, a cup of tea on the side table. Warm home aesthetic.

What NB2 produces: Getting there — there's a person, a scene, and a mood. The Image Search for "tall windows natural light" and the scene description will anchor NB2 well. But the photographic style is still unspecified, and the "premium" quality of the headphones isn't conveyed.

Problem: No camera direction, no lighting specifics, no color grade. NB2 will make its own choices — they might not match your brand.

Iteration 4: Add Camera, Lighting, and Style

A young woman in her late 20s with short natural hair sits in a beautifully designed Scandinavian-style living room, wearing large matte black over-ear headphones. Eyes closed, a peaceful expression. Shot on an 85mm f/1.8 lens, medium close-up framing. Soft golden afternoon light from floor-to-ceiling windows behind her, creating a warm backlit halo effect. A minimalist coffee table with a white ceramic mug slightly out of focus in the foreground. Kodak Portra 400 color rendering, slightly warm film tones. Lifestyle photography aesthetic for a premium consumer electronics brand.

What NB2 produces: This is close to production-ready. Specific enough to have a distinct look. The 85mm lens creates portrait compression, the backlit halo makes the scene feel premium, and Kodak Portra gives the color a filmic quality.

Problem: The headphones are described as generic "matte black." In a real campaign, you'd want visual accuracy. And the exclusions list is missing.

Iteration 5: Final Production-Ready Version

A young woman in her late 20s with short natural hair and warm brown skin sits in a beautifully designed Scandinavian-style living room, wearing the Sony WH-1000XM5 headphones (matte black). Eyes closed, a peaceful, fully immersed expression, slight smile. Shot on an 85mm f/1.8 lens, medium close-up framing. Soft golden late-afternoon light from floor-to-ceiling windows behind her creates a warm backlit halo effect. A minimalist wooden coffee table with a white ceramic mug sits slightly out of focus in the foreground right. Kodak Portra 400 color rendering, slightly warm film tones, subtle grain. Lifestyle photography aesthetic for a premium consumer electronics brand campaign. 3:2 aspect ratio, 4K resolution. Avoid: blurry faces, extra fingers, artifacting, oversaturated colors, generic stock photo feeling.

The progression in 5 steps:

1. Vague concept
2. Add mood and context
3. Ground in a specific scene with character
4. Add camera, lighting, and style vocabulary
5. Add product accuracy (Image Search), specific details, exclusions, resolution

Each iteration adds one layer of control. In a real workflow, you'd also run iteration 4 at 0.5K to confirm the composition, then move to 4K only for the finalist.

16. NB2 vs. Pro Decision Flowchart

Use this flowchart when you're not sure which model to use. Answer each question and follow the path.

START: What type of image are you generating?
│
├─ Do you need an ultra-wide or ultra-tall aspect ratio?
│   (8:1, 4:1, 1:4, 1:8)
│   ├─ YES → USE NB2. Only NB2 supports these ratios. ✅
│   └─ NO → Continue ↓
│
├─ Does your image reference a specific real-world location,
│   product, brand, or landmark that needs visual accuracy?
│   ├─ YES → STRONGLY CONSIDER NB2. Its Image Search retrieves
│   │         real visual references. (Pro can't do this.) ✅
│   └─ NO → Continue ↓
│
├─ Are you in an exploration or iteration phase?
│   (Generating multiple concepts, testing compositions, A/B testing)
│   ├─ YES → USE NB2. It's ~2.5x faster and ~70-80% cheaper.
│   │         Save Pro for final polish. ✅
│   └─ NO → Continue ↓
│
├─ Are you generating high-volume content?
│   (10+ images in a batch, social series, product catalogues)
│   ├─ YES → USE NB2. Cost and speed at scale make it the
│   │         clear choice for volume work. ✅
│   └─ NO → Continue ↓
│
├─ Does this image need the maximum possible aesthetic quality?
│   (Hero shot for a flagship campaign, premium print, fine art output)
│   ├─ YES → Continue ↓
│   └─ NO → USE NB2. For most content, NB2's quality
│             is more than sufficient. ✅
│
├─ Is budget a concern for this project?
│   ├─ YES → USE NB2. At ~$0.039 vs. up to $0.24 per image,
│   │         the savings are significant at any scale. ✅
│   └─ NO → Continue ↓
│
├─ Does this image require complex compositional reasoning?
│   (Intricate spatial relationships, multi-element symbolism,
│    unusual interpretive challenges)
│   ├─ YES → CONSIDER NB2 PRO. Its thinking mode handles
│   │         complex compositional reasoning better. ✅
│   └─ NO → Continue ↓
│
└─ Is this a single premium output where quality is the only criterion?
    ├─ YES → USE NANO BANANA PRO. This is its sweet spot. ✅
    └─ NO → USE NB2. It's the right default for professional work.

Summary of default positions:

• NB2 is the right choice for ~80% of professional creative work
• Pro is worth the premium for single flagship outputs where quality is paramount
• NB2 is always the right choice when you need ultra-wide formats or real-world visual accuracy

17. Cheat Sheet

Quick reference for everything in this guide. Print it out, pin it up.

Model at a Glance

Exclusive Features (NB2 Only)

• Ultra-wide ratios: 8:1, 4:1, 1:4, 1:8
• Image Search: Real-time visual reference lookup for landmarks, brands, products

Aspect Ratio Quick Reference

Resolution Workflow

0.5K → Draft/explore (cheapest, fastest)
1K   → Refine / social media
2K   → Digital production standard
4K   → Print / hero outputs only

The Prompt Formula

[Subject + details] + [Action/state] + [Environment] + 
[Composition/camera] + [Lighting/mood] + [Style] + 
[Avoid: ...]

Camera Vocabulary Pocket Reference

Lighting Pocket Reference

Film Stock Quick Picks

Text in Images

1. Always use "quotation marks" around text to render
2. Specify font style: "bold sans-serif", "hand-painted", "neon cursive"
3. For complex/critical copy: generate text first, then embed in image

The Quality Control Checklist

Before submitting any prompt, ask:

• [ ] Did I write in full sentences (not keywords)?
• [ ] Did I specify subject, environment, lighting, AND style?
• [ ] Did I include camera/lens direction?
• [ ] Is the aspect ratio specified?
• [ ] For text: did I use quotation marks?
• [ ] For real locations/brands: did I name them specifically?
• [ ] Did I add an exclusions list for photorealistic images?
• [ ] Am I at the right resolution for this stage (draft vs. final)?

NB2 vs. Pro in One Line

Guide compiled and expanded from: Google Cloud Blog — Ultimate Prompting Guide for Nano Banana, TechCrunch — Google launches Nano Banana 2, Android Central — Google announces Nano Banana 2, Banana AI — Nano Banana 2 Features Review, Freepik Blog — Nano Banana 2. Last updated: April 2026.