Nano Banana Pro — Complete Prompting Guide

A Human-Friendly Reference for Creative Professionals and Enthusiasts

Welcome: What Is Nano Banana Pro and Why Should You Care?

Imagine having a world-class visual artist on call — one who can interpret a vague creative direction, fill in sensible details, understand lighting physics, and deliver print-quality results in about ten seconds. That's the closest analogy to what Nano Banana Pro offers.

Nano Banana Pro is Google DeepMind's flagship image generation and editing model, built on the Gemini 3 Pro Image architecture and released on November 20, 2025. It sits at the top of the image generation landscape for a specific reason: it thinks before it draws. Unlike most image generators that treat your prompt like a checklist of keywords to activate, Nano Banana Pro runs chain-of-thought reasoning over your words before producing a single pixel. It weighs physics, composition, spatial relationships, and coherence the way a skilled illustrator would before picking up a brush.

For creative professionals — photographers, brand designers, art directors, illustrators, game designers, marketers — this is a meaningful shift. You no longer need to speak the machine's language (keyword soup, arcane tags, magic numbers). You speak like a human, like a Creative Director briefing a talented artist. Full sentences with clear intent. Purpose-driven descriptions. The more clearly you can communicate what you want and why, the better your results.

This guide is your mentor for that conversation. It covers everything from the raw technical specs (if you need them) to deep creative strategies, copy-paste-ready example prompts, and a workshop that shows you how a great prompt is built from scratch. Whether you've never used an image generator or you're migrating from another tool, you'll find what you need here.

Quick Start — Jump Right In

Not ready to read the whole guide yet? Here are three copy-paste-ready prompts you can use immediately. Each one demonstrates a different strength of the model.

Quick Start Prompt 1 — Product Hero Shot

A high-end commercial product shot of a matte ceramic candle in pale sage green, sitting on a smooth white marble slab. Three-point softbox lighting, soft warm fill from the left. A single dried eucalyptus sprig rests beside the candle. Shallow depth of field (f/2.8). The label reads "CALM" in thin, elegant Garamond italic. Shot on medium-format Hasselblad. 1:1 aspect ratio, 2K resolution.

Quick Start Prompt 2 — Editorial Portrait

A fashion editorial portrait of a tall man in his 40s with close-cropped grey hair, wearing a structured midnight-blue overcoat over a black turtleneck. He stands on a fog-covered cobblestone street in Edinburgh at blue hour, looking off-frame to the right with quiet intensity. Low angle, 50mm lens, f/1.8, Kodak Portra 800 film grain. 3:4 aspect ratio.

Quick Start Prompt 3 — Illustrated Scene

A lush watercolor illustration of a tiny cottage built inside the hollow of an ancient oak tree, with warm amber light glowing from the windows. Autumn leaves drift past. A stone path winds from the door into a misty forest. Loose, expressive brushwork, deep saturation, reminiscent of a high-end children's book illustration by Shaun Tan. 3:4 aspect ratio, 2K resolution.

Once you're comfortable with the basics, come back and read the full guide — there's a lot of depth to explore.

Table of Contents

1. Model Identity
2. Technical Specifications
3. Core Architecture Insight — Why This Model Is Different
4. Golden Rules of Prompting
5. Prompting Frameworks
6. Text Rendering
7. Creative Director Controls
8. Character and Brand Consistency
9. Multi-Image Fusion
10. Aspect Ratio and Resolution
11. The Grid Trick (Advanced)
12. Nano Banana Pro vs. Nano Banana 2 — When to Use Which
13. Common Pitfalls
14. Production-Ready Example Prompts
15. Common Mistakes → Better Versions
16. Prompt Building Workshop
17. Cheat Sheet — Quick Reference

1. Model Identity

Every image Nano Banana Pro generates is invisibly watermarked using Google's SynthID technology and includes C2PA Content Credentials, an industry-standard metadata standard that declares AI involvement in the creation. This is baked in — you can't opt out, and it's a deliberate design choice for responsible AI deployment.

2. Technical Specifications

What These Numbers Mean in Practice

On resolution: 2K (2048×2048) is the baseline — it's 4 megapixels, which is more than enough for social media, web use, and most print up to A4/letter size. 4K (4096×4096) is 16 megapixels — suitable for large-format printing, billboard mock-ups, and production assets where fine detail is critical.

On reference images: You can feed up to 14 images at once in a workflow. Up to 6 of those will be processed with high fidelity (meaning the model closely preserves their visual identity). The remaining 8 serve as contextual references. Think of high-fidelity slots as your "anchor" images (brand logo, hero character, product shape) and contextual slots as inspiration or mood references.

On cost: At $0.134 per 1K/2K image, a batch of 50 exploratory images costs under $7. 4K images at $0.24 are still cheap compared to a half-hour of a professional retoucher's time. For high-volume work, Nano Banana 2 is more economical (see section 12).

3. Core Architecture Insight

Why This Model Is Different — and Why It Changes How You Write Prompts

Most image generators process your prompt as a bag of words — they activate concepts probabilistically without fully "understanding" what you mean. Nano Banana Pro works differently.

It is a "thinking" image model. Before generating a single pixel, it performs chain-of-thought reasoning over your prompt. Internally, it asks itself questions like: What is the subject doing? What are the lighting physics of this scene? Does the perspective make spatial sense? What style is implied by these details?

This has three important practical implications:

1. It interprets intent, not just keywords. If you write "a quiet morning coffee scene," the model infers ambient light, a relaxed mood, a warm color palette — without you specifying each one.
2. It self-augments underspecified prompts. If you leave something vague, the model fills in a sensible default rather than producing garbage or randomness. (This is great for getting started, but you'll usually want more control — which is what this guide gives you.)
3. Full sentences dramatically outperform keyword lists. Because the model is reasoning over language, coherent prose gives it far more to work with than fragmented tags. "warm, golden, coffee, bokeh, 4k, cozy" tells the model much less than "A ceramic mug of steaming coffee on a worn wooden table, photographed in warm morning light with soft window bokeh, evoking a slow, quiet Sunday." The second prompt has intent — the model can reason from it.

4. Golden Rules of Prompting

These seven rules distill everything that matters about writing effective prompts for Nano Banana Pro. Master these before worrying about any advanced technique.

Rule 1 — Use Natural Language, Not Tag Soup

Write full sentences with proper grammar. The model reasons over language, so coherent prose gives it far more context than fragmented keywords.

Why this works: The thinking engine processes narrative context. A sentence communicates relationships between elements (the dog is in the park, playing), mood (sunlit), and intent (photorealistic). Tags give it isolated concepts with no connective tissue.

Rule 2 — Use Positive Framing

Describe what you want in the image, not what you want to exclude.

Why this works: The model generates from positive descriptions. When you say "no cars," it still has to imagine cars to know what to avoid — and sometimes loses that battle. Describing the desired state positively is unambiguous.

Rule 3 — Edit, Don't Re-Roll

If a generated image is 80% correct, don't start over. Conversationally request specific changes within the same session.

Example follow-up: "That's great, but change the lighting to sunset and make the text neon blue."

Why this works: Nano Banana Pro maintains session context. Your original generation becomes a reference point the model knows intimately — small edits leverage that context rather than rebuilding from nothing. Re-rolling from scratch means re-winning a lottery you'd already mostly won.

Rule 4 — Be Specific and Descriptive

Vague subjects produce generic images. The more specific your description, the more distinctive and intentional the result.

Why this works: Specific details don't just add information — they establish a voice for the image. The more confident and particular your description, the more the model can lean into that vision rather than defaulting to the average of everything it has seen.

Rule 5 — Provide Context (The "Why")

Include the purpose of the image when relevant.

Why this works: Purpose changes compositional priorities. "For a children's book illustration" implies soft edges, approachable colors, and readable forms at small size. "For a billboard mockup" implies bold contrast and simple focus. Context helps the model make the right trade-offs.

Rule 6 — Control the Camera

Think like a photographer or cinematographer. Use specific, technical camera language to precisely specify your composition.

Useful terms:

• Shot type: "close-up", "medium shot", "full body", "aerial view", "worm's-eye view"
• Lens: "85mm portrait lens", "24mm wide-angle", "macro lens", "200mm telephoto compression"
• Aperture / depth of field: "shallow depth of field (f/1.4)", "deep focus (f/11)", "soft bokeh background"
• Angle: "low angle", "high angle", "dutch tilt", "straight-on"

Why this works: Camera parameters aren't just technical — they're compositional and emotional choices. A low-angle f/1.4 shot of a person feels powerful and intimate. An aerial f/8 shot of the same person feels observational and detached. Specifying these controls puts that decision in your hands.

Rule 7 — Iterate Conversationally

Your first generation is rarely your final result — and it doesn't need to be. Refine with follow-up prompts in the same session, building toward your vision incrementally.

5. Prompting Frameworks

These are the five core "shapes" a prompt can take, depending on what you're doing and what inputs you have. Each has a formula to follow and a worked example.

5.1 Text-to-Image (No Reference Images)

This is the most common use case: you have words, and you want an image.

Formula:

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

Why This Works: This five-part structure covers the minimum viable information for a well-specified image. Subject and Action answer who/what and what are they doing? Location/Context answers where? Composition answers how are we looking at this? Style answers what does it feel like? Together they create a complete creative brief — nothing is left to default.

Example:

[Subject] A striking fashion model wearing a tailored brown dress, sleek boots, and holding a structured handbag.
[Action] Posing with a confident, statuesque stance, slightly turned.
[Location/Context] A seamless, deep cherry red studio backdrop.
[Composition] Medium-full shot, center-framed.
[Style] Fashion magazine style editorial, shot on medium-format analog film, pronounced grain, high saturation, cinematic lighting effect.

In practice, you write these as a flowing paragraph:

A striking fashion model wearing a tailored brown dress, sleek boots, and holding a structured handbag. She poses with a confident, statuesque stance, slightly turned toward camera, against a seamless, deep cherry red studio backdrop. Medium-full shot, center-framed. Fashion magazine editorial style, shot on medium-format analog film with pronounced grain, high saturation, and cinematic lighting.

Both formats work. The paragraph flows more naturally — the formula version is useful when learning or debugging a prompt that isn't working.

5.2 Multimodal Generation (With Reference Images)

You have one or more reference images and want to use them as inputs.

Formula:

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

Why This Works: This structure tells the model what you're providing, how to use it, and what to produce. The "relationship instruction" is the critical middle piece — without it, the model has to guess how much to borrow from each reference. Being explicit about roles ("use this as the texture", "use this as the structure") removes ambiguity and dramatically improves output consistency.

Example:

Using the attached napkin sketch as the structure and the attached fabric sample as the texture [References], transform this into a high-fidelity 3D armchair render [Relationship]. Place it in a sun-drenched, minimalist living room [New Scenario].

Reference role vocabulary: Here are useful phrases for defining reference relationships:

• "Use Image A as the character's visual identity" — for locking people
• "Use Image B as the color palette reference" — for mood/color
• "Use Image C as the structural blueprint" — for shapes/layout
• "Use Image D as the background environment" — for settings
• "Use Image E as the style reference" — for artistic style

5.3 Image Editing (Conversational, No New References)

You want to modify an image the model just generated, using natural language. Nano Banana Pro uses semantic masking — you define what to change and what to keep using words, not selection tools.

Formula (implicit):

[What to change] + [How to change it] + [What to keep]

Why This Works: The model already "knows" everything about the image it generated. You don't need to re-describe the whole scene — just describe the delta. Being explicit about what to preserve is just as important as describing what to change, because it gives the model permission to leave the rest alone.

Examples:

• "Remove the man from the photo" — object removal
• "Change the background to a neon-lit Tokyo street at night, keep the subject identical" — background swap with explicit preservation instruction
• "Make it nighttime" — high-level scene transformation (model infers lighting, sky, shadows, reflections)
• "Replace the red jacket with a tailored charcoal blazer, keep everything else the same" — targeted clothing change

5.4 Composition and Style Transfer (With New References)

Two specialized use cases: combining elements from different images, or translating an image's style.

Adding elements: Upload a base image and an object image. Instruct the model to place the object into the scene. Be specific about placement, scale, lighting match, and integration.

Take the sneaker from Image A and place it on the polished wooden floor in Image B, slightly left of center, with a matching soft shadow underneath and the same warm ambient lighting from the scene.

Style transfer: Upload a photograph and request its recreation in a different artistic style. The model preserves the content while reimagining the rendering.

Recreate this photograph of the modern city street as a Van Gogh-style oil painting — thick, expressive impasto brushstrokes, swirling skies in deep navy and gold, vivid saturated color throughout.

Why This Works: The model's multimodal reasoning can separately process "what is in this image" (content) and "how is it rendered" (style) and recombine them. Be specific about which aspects of the style reference you want applied — "brushwork", "color palette", "line character", "texture" — to get precise results.

5.5 Real-Time Web Search Integration

Nano Banana Pro has live web search access, which means you can ask it to look up real-world information and incorporate it into an image.

Formula:

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

Why This Works: Grounding an image in real data makes it immediately relevant and non-generic. A data visualization prompt that includes live market cap numbers is more accurate and compelling than one made up from imagination. The three-part formula ensures you get the right data, processed the right way, rendered visually.

Example:

Search for the current weather and date in San Francisco. Use this data to create a miniature city-in-a-cup concept, modifying the scene to reflect the real weather conditions (e.g., if raining, show grey skies and wet surfaces). Embed it within a realistic, modern smartphone UI mockup.

More use cases:

• Real stock prices → custom financial infographic
• Today's top news headline → editorial illustration
• Current sports standings → trophy visualization
• Live cryptocurrency prices → dashboard mock-up

6. Text Rendering

Nano Banana Pro has best-in-class text rendering among image generation models, supporting accurate text across multiple languages including Korean, Japanese, Chinese, Arabic, and European scripts. Here's how to use it effectively.

The Five Rules of Text in Prompts

1. Enclose desired text in quotation marks. Always. "Happy Birthday", "URBAN EXPLORER", "Sale Ends Sunday". Without quotes, the model may guess at text content or render something approximate.
2. Specify typography style. Don't just say "text" — say what kind. Examples:

• "bold, white, sans-serif font"
• "Century Gothic, 12px, letter-spaced"
• "neon cursive signage, glowing pink"
• "hand-lettered chalk style"
• "embossed gold serif, Art Deco style"

1. Multilingual support is reliable. Korean, Japanese, Chinese, Arabic, and major European languages render accurately. Specify the language explicitly if you want non-English text: "render the tagline in Korean".
2. Text-first workflow for critical copy. If text content is important (a logo, an ad headline, a poster title), finalize the text first in conversation, then request the image. This reduces the chance of subtle rendering errors in long or complex copy.
3. Translation and localization. You can request the model to render text in one language and also provide translated versions: "render the headline in English, then below it in Arabic (right-to-left), then in Korean".

Example: Multi-Typography Product Shot

A high-end, glossy commercial beauty shot of a sleek, minimalist nude-colored face moisturizer jar. Next to the product, render three lines of text: top line "GLOW" in a flowing, elegant Brush Script font; middle line "10% OFF" in a heavy, blocky Impact font; bottom line "Your First Order" in thin, minimalist Century Gothic font. Then translate the text into Korean and Arabic.

7. Creative Director Controls

This section is your vocabulary reference. These are the specific, technical terms that give you fine-grained control over the look and feel of generated images.

7.1 Lighting

Lighting is the single highest-leverage control in photography and image generation. Getting it right transforms a competent image into a striking one.

7.2 Camera, Lens, and Focus

Think of camera controls as a hierarchy: first choose the shot type (how much of the scene you see), then the lens (how it's distorted/compressed), then the focus (what's sharp).

Camera hardware choices:

• "GoPro" → immersive, wide, slightly distorted — great for action and immersive POV
• "Fujifilm X-series" → authentic, slightly warm color science — great for lifestyle and street
• "Disposable camera" → raw, high-flash, nostalgic — great for editorial and lo-fi aesthetics
• "Hasselblad medium format" → maximum detail, rich tonal range — great for product and high-end portrait
• "Large-format film camera" → extreme detail, slight perspective distortion — great for fine art

Lens and focal length:

• "24mm wide-angle" → expansive environments, slight distortion
• "35mm" → natural perspective, documentary feel
• "50mm" → closest to human eye — neutral, clean
• "85mm portrait lens" → flattering compression, slight background separation
• "135mm" → strong compression, intimate feel at distance
• "200mm telephoto" → heavy compression, subjects feel stacked/close
• "macro lens" → extreme close-up with sharp detail across a tiny plane

Depth of field:

• "f/1.2" or "f/1.4" → extremely shallow — subject sharp, everything else melts into bokeh
• "f/2.8" → shallow — standard portrait/product shallow focus
• "f/8" → moderate depth — most of the scene in focus
• "f/11–f/16" → deep focus — everything sharp, landscape photography standard

7.3 Color Grading and Film Stock

Color grading is mood in visual shorthand. Use these references for precise emotional and aesthetic targeting:

7.4 Materiality and Texture

Be explicit and specific about materials. This is especially important for product photography, fashion, and architectural renders.

Material vocabulary:

• Fabric: "navy blue tweed", "raw silk dupioni", "weathered canvas", "cashmere ribbed knit", "sheer organza", "distressed denim"
• Metal: "brushed steel", "hammered copper", "anodized matte black aluminum", "polished 18k gold", "aged patina bronze"
• Ceramic and glass: "matte stoneware glaze", "handmade speckled ceramic", "frosted borosilicate glass", "iridescent glaze"
• Natural: "rough-hewn reclaimed oak", "smooth Carrara marble", "woven rattan", "polished obsidian"
• Synthetic: "soft-touch matte rubber", "high-gloss UV lacquer", "carbon fiber weave", "translucent TPU"
• Fantasy/textural: "ornate elven plate armor, etched with silver leaf patterns", "liquid mercury surface", "crumpled velvet", "bioluminescent skin"

8. Character and Brand Consistency

One of Nano Banana Pro's standout capabilities is maintaining visual coherence across multiple images — for both characters and brand assets.

Character Identity Locking (Multi-Image Workflows)

When you need the same person to appear across multiple images:

Step 1 — Prepare your reference images. Use clear reference photos: front-facing or slight 3/4 angle, good even lighting, minimal background clutter. Multiple references (different angles, lighting conditions) improve consistency.

Step 2 — Be explicit about what to lock. In your prompt: "Keep the person's facial features exactly the same as Image 1." Don't leave this implicit.

Step 3 — Describe the changes, anchor the identity. Change emotion, pose, clothing, and scene. Keep the physical identity description present in the prompt as a reinforcement anchor.

Step 4 — Use consistency limits wisely. The model maintains visual coherence for up to 5 people in a single workflow. For larger casts, split into separate sessions.

Example workflow prompt:

Using the attached reference image as the character: Generate a series of 4 images showing this character in different scenarios. Image 1: Walking through a rainy Tokyo street at night with neon reflections. Image 2: Sitting in a sunlit Parisian café reading a book. Image 3: Standing on a cliff overlooking the ocean at sunset. Image 4: Working at a desk in a modern, minimalist apartment. Keep the person's facial features, body proportions, and hair exactly the same as the reference. Vary only the clothing and setting. Photorealistic style, 2K resolution, 1:1 aspect ratio.

Brand Consistency

For brand work — campaigns, product visualizations, social media series:

1. Upload brand assets as reference images: logos, color swatches, product photography, packaging design.
2. Anchor explicitly: "Maintain the brand's color palette and typography style from the reference."
3. Brand dressing: You can drape patterns, logos, and artwork onto 3D objects while preserving natural lighting and texture — useful for product mock-ups and brand activations.

Example: Upload a brand's logo + color palette + product reference, then:

Create a lifestyle photo showing the product (Image A) being used outdoors in summer, incorporating the brand's coral and sand color palette (Image B). The logo from Image C should appear subtly on the product label. Warm golden hour lighting, natural and editorial feel.

9. Multi-Image Fusion

Nano Banana Pro can blend up to 14 separate images into a cohesive composition. This is one of its most powerful capabilities for complex production work.

The cardinal rule: define the role of every image you provide.

Without clear role assignments, the model must guess what each image is for, which leads to unpredictable blending. With clear roles, you're directing a precise creative process.

Role-definition template:

Use Image A for [role]. Use Image B for [role]. Use Image C for [role].

Example roles:

• "Use Image A for the character's pose and expression"
• "Use Image B for the art style and color palette"
• "Use Image C for the background environment"
• "Use Image D for the product shape and proportions"
• "Use Image E as a texture/material reference"

Fashion/apparel example:

Combine these images into one appropriately arranged cinematic image in 16:9 format and change the dress on the mannequin to the dress shown in Image B.

Complex production example:

Use Image A for the character's pose, Image B for the art style (painterly, Studio Ghibli-influenced), and Image C for the forest background environment. Compose a full scene with the character in the foreground, the Ghibli-style rendering applied throughout, and the forest from Image C visible behind. Soft dappled light, 4:3 aspect ratio, 2K resolution.

10. Aspect Ratio and Resolution

Aspect Ratio Reference

How to specify in prompts:

• "A 9:16 vertical poster for Instagram Stories"
• "A cinematic 21:9 wide shot"
• "Output at 4K resolution"
• "Generate at 2K, 3:4 portrait format"

Resolution Guide

Generating Variations in One Prompt

For rapid exploration, request multiple variations in a single generation:

• "Generate three distinct variations of this product mockup"
• "Show four different color palette options for this scene"
• "Create two versions: one with warm lighting, one with cool"

11. The Grid Trick (Advanced)

Here's a technique that gets you more value per generation — and it's unique to Nano Banana Pro's architecture.

The setup: Nano Banana Pro generates at 2K baseline (4 megapixels total). Each "sub-image" in a 2×2 grid is approximately 1 megapixel — equivalent to the full output of the original Nano Banana model. You can request a 2×2 grid in a single generation and get four distinct images.

Why it's useful:

• You get four variations in one ~10-second generation
• Each sub-image can be meaningfully different because of the model's autoregressive generation process (each panel is generated sequentially, with slight variation)
• Excellent for rapid concept exploration — pick the best, then iterate at full resolution

How to use it:

Generate a 2×2 grid of four distinct variations of [your concept]. Each panel should explore a different [lighting condition / color palette / composition / time of day]. 2K resolution, 1:1 aspect ratio.

Example:

Generate a 2×2 grid of four variations of a coffee shop window scene. Panel 1: Morning with golden sunrise light. Panel 2: Afternoon with harsh midday shadows. Panel 3: Evening with warm interior light glowing out. Panel 4: Night with only streetlight and candlelight. 2K resolution.

12. Nano Banana Pro vs. Nano Banana 2 — When to Use Which

Nano Banana Pro is not always the right choice. Here's a clear guide for making the decision.

The practical rule: Use Nano Banana Pro when the image is going somewhere important — a client presentation, a print campaign, a hero visual on a website. Use Nano Banana 2 when you're exploring, iterating at speed, or need ratios that Pro doesn't support.

13. Common Pitfalls

14. Production-Ready Example Prompts

This is the most expanded section of the guide. For each category, you'll find the original production example plus at least three additional prompts, each with a brief note explaining the key choices made.

14.1 Product Photography

Original Example:

A high-end commercial product shot of a matte black wireless headphone on a polished marble surface. Three-point softbox lighting with soft reflections. The headphone cable is coiled naturally. Shallow depth of field (f/2.8). The text "AURA PRO" appears in clean, white Helvetica Neue Light font at the top of the frame. Shot on a medium-format Hasselblad. 16:9 aspect ratio, 4K resolution.

Skincare — Minimalist Hero Shot

A commercial hero shot of a frosted glass serum bottle with a gold dropper, centered on a smooth white travertine surface. Single large window creating soft, directional natural light from the left with a clean shadow. A single dried white wildflower rests beside the bottle, slightly out of focus. Shallow depth of field (f/2.0). Shot on Hasselblad medium format. 1:1 aspect ratio, 4K resolution.

Key choices: travertine over standard marble adds texture without competing; the dried flower adds organic warmth without color noise; f/2.0 keeps the product sharp while softening the surface's texture.

Spirits — Dark & Moody Bar Shot

A dramatic commercial product shot of a bottle of single malt Scotch whisky on a dark, oiled mahogany bar surface. Chiaroscuro lighting — a single warm overhead light source creates a strong pool of amber light around the bottle, with deep shadows falling away. Two heavy crystal tumblers, one with ice, flank the bottle. Slight condensation on the glass. Shot on 85mm, f/2.8. 3:2 aspect ratio, 2K resolution.

Key choices: Chiaroscuro lighting matches the "moody, premium" brand positioning of whisky; mahogany and crystal communicate heritage; condensation adds tactile realism.

Tech — Floating Perspective Shot

A product photography mock-up of a slim wireless charging pad in matte white, photographed from slightly above and to the side — a 45-degree overhead angle with a soft floating shadow beneath it. Pure white background. A slim smartphone sits on the pad, screen glowing softly. Even, diffused studio lighting with no harsh highlights. Clean, Apple-style product aesthetic. 1:1 aspect ratio, 2K resolution.

Key choices: 45-degree overhead angle creates the "floating product" look standard in tech marketing; the soft shadow below grounds the product without a physical surface; "Apple-style" aesthetic is a well-understood reference the model interprets correctly.

Candle — Lifestyle Flatlay

A lifestyle product flatlay of three soy candles in minimalist concrete vessels — sage green, warm terracotta, and bone white — arranged in a loose triangle formation on a rough linen textile. Scattered dried botanicals (lavender, rosemary, dried orange slices) fill the negative space. Shot from directly overhead. Diffused natural window light, soft and shadowless. 1:1 aspect ratio, 2K resolution. Editorial, editorial, natural aesthetic.

Key choices: three color SKUs visible in one shot; flatlay maximizes legibility for social media; botanicals reinforce "natural/artisan" positioning without being cluttered.

14.2 Editorial Fashion

Original Example:

A fashion editorial portrait of a young woman with short silver hair, wearing an oversized deconstructed blazer in charcoal wool over a sheer white blouse. She leans against a concrete pillar in an abandoned industrial space. Golden hour sunlight streams through broken windows, casting warm geometric shadows. Low angle shot, 35mm lens, f/1.4. Film grain reminiscent of Kodak Portra 800. 3:4 aspect ratio.

Street Fashion — Urban Realism

A street fashion editorial of a young man, early 20s, with closely cropped natural hair, wearing an oversized vintage track jacket in forest green and a pair of wide-leg cargo trousers, white Air Force 1s. He stands mid-stride on a wet East London street at dusk, graffiti-covered railway arches behind him. Ambient blue-grey light, specular reflections on the wet tarmac. Shot on a 35mm lens, f/2.0. Fujifilm Pro 400H color grading. 3:4 aspect ratio, 2K resolution.

Key choices: "mid-stride" creates energy and naturalism; wet tarmac adds visual richness; railway arches set a specific, credible subculture context; Fujifilm Pro 400H gives soft, slightly desaturated fashion-editorial tones.

High Fashion — Studio Abstraction

An avant-garde fashion editorial. A woman in her 30s, statuesque, wearing a sculptural pleated dress in iridescent white silk that catches the light in shifting colors. She stands against a seamless dark charcoal backdrop, facing directly into camera with a completely neutral expression. A single hard light source from slightly above and to the right creates sharp, dramatic shadows that emphasize the dress's structure. Shot on 50mm, f/5.6 for slightly deeper focus. Vogue Italia editorial aesthetic. 2:3 aspect ratio, 4K resolution.

Key choices: iridescent silk rewards 4K resolution where color shift detail matters; neutral expression keeps attention on the garment; f/5.6 slightly deeper focus is unconventional here but keeps the dress's texture sharp throughout.

Resort Wear — Location Editorial

A sun-drenched resort wear editorial of two women, both early 30s, wearing flowing linen co-ord sets — one in warm saffron, one in pale aqua. They stand on a white-painted terrace overlooking a deep blue Mediterranean sea, laughing in natural conversation. Photographed from a medium distance, natural backlighting at golden hour creating rim light around their hair. Shot on a 50mm lens, f/2.8. Fujifilm Velvia-influenced color for saturated blues and yellows. 3:2 aspect ratio, 2K resolution.

Key choices: two characters with complementary colors creates immediate visual interest; "laughing in natural conversation" is specific and creates authentic movement; Velvia color reference maximizes the Mediterranean palette.

Menswear — Autumn Campaign

A cinematic menswear editorial portrait of a man in his 50s with a weathered, distinguished face and silver stubble, wearing a waxed canvas hunting jacket in forest green over a chunky oatmeal ribbed knit, dark brown corduroy trousers, and worn leather boots. He stands at the edge of a misty autumn forest in the Scottish Highlands, looking out across a valley. Early morning, flat grey light. Shot on 135mm telephoto, f/2.8, creating heavy atmospheric compression between the subject and the distant hills. Muted, earthy Kodak Portra 400 palette. 3:2 aspect ratio, 2K resolution.

Key choices: 135mm telephoto compression pulls the distant hills closer, creating the "man standing against landscape" drama; waxed canvas and corduroy communicate authentic outdoors heritage; flat grey light is deliberate — the muted Scottish atmosphere is the aesthetic.

14.3 Character Consistency

Original Example:

Using the attached reference image as the character: Generate a series of 4 images showing this character in different scenarios. Image 1: Walking through a rainy Tokyo street at night with neon reflections. Image 2: Sitting in a sunlit Parisian café reading a book. Image 3: Standing on a cliff overlooking the ocean at sunset. Image 4: Working at a desk in a modern, minimalist apartment. Keep the person's facial features, body proportions, and hair exactly the same as the reference. Vary only the clothing and setting. Photorealistic style, 2K resolution, 1:1 aspect ratio.

Children's Book Character — Across Scenes

Using Image A as the visual reference for the character — a small raccoon in a red scarf and round spectacles — create three illustrations of this character in different moments: Scene 1: Sitting under a large mushroom in a rainy forest, reading a tiny book. Scene 2: Looking up in wonder at a night sky full of stars from a hilltop. Scene 3: Stirring a bubbling pot in a cozy underground kitchen, jars of magical ingredients on shelves behind. Maintain the raccoon's exact appearance, scarf color, and spectacle style across all three. Watercolor illustration style, warm palette, children's book aesthetic reminiscent of Beatrix Potter. 3:4 aspect ratio, 2K resolution.

Key choices: precise visual anchors (red scarf, round spectacles) give the model strong identity cues; three complete scene descriptions give distinct context; Beatrix Potter reference establishes the illustration tradition precisely.

Game Character — Environment Variations

Using the reference image (Image A) as the definitive design for the character — a female mercenary, mid-30s, with short-cropped auburn hair, a bionic left arm with exposed cable details, and a worn leather duster coat — generate three environment shots: Scene 1: Emerging from the shadows of a rain-soaked cyberpunk alley, neon signs overhead. Scene 2: Standing on a sun-bleached desert highway at high noon, dust devil in the background. Scene 3: Crouched in the ruins of an overgrown abandoned office, shafts of light through broken windows. Keep facial features, bionic arm design, and duster coat identical. Vary lighting, environment, and expression. Photorealistic digital art style. 16:9 aspect ratio, 2K resolution.

Key choices: the bionic arm is a distinguishing visual anchor that helps the model maintain identity consistency; three dramatically different lighting environments (rain/neon, high noon, ruin light) test and demonstrate the model's consistency range.

Brand Mascot — Marketing Use Cases

Using Image A as the definitive reference for the brand mascot — a friendly, plump golden bear wearing a white chef's apron with the letter "B" embroidered in blue — create four images showing the mascot in different brand contexts: Image 1: Standing behind a kitchen counter, holding up a plate of fresh pastries, waving. Image 2: Delivering a box labeled "FRESH TODAY" on a bicycle, city street behind. Image 3: Working at a laptop in a cheerful modern office, coffee nearby. Image 4: On a bright outdoor market stall surrounded by produce. Keep the bear's appearance, apron design, and embroidered "B" identical in all four. Friendly, colorful illustration style, white backgrounds with warm accent colors. 1:1 aspect ratio, 2K resolution.

Key choices: "B" embroidered in blue is a specific identity anchor; four marketing contexts cover digital, delivery, office, and market use cases; white backgrounds with warm accents are practical for compositing into other materials.

Portrait Series — Documentary Style

Using the attached portrait reference (Image A) as the character: Create three documentary-style portrait images of this person in different occupational contexts. Image 1: Working at a busy fish market at dawn, apron on, surrounded by ice and fresh catch. Image 2: Seated in a quiet library archive, surrounded by old maps and documents, wearing reading glasses. Image 3: Standing at the entrance of a small auto repair garage, grease-stained gloves in hand, calm and proud expression. In each image, maintain the person's exact facial features, age, and general build from the reference. Change clothing, setting, expression to match the occupation. Photorealistic, warm natural light, human documentary photography style. 3:4 aspect ratio, 2K resolution.

Key choices: "calm and proud expression" is more specific than "happy" or "smiling" — documentary photography prizes dignity; the variety of occupational contexts demonstrates character range for casting-style workflows.

14.4 Infographic

Original Example:

Search for the current top 5 most valuable companies by market capitalization. Create a clean, modern infographic displaying each company's name, logo, and market cap value. Use a dark navy background with white text in Inter font. Arrange as a horizontal bar chart. Style: minimal data visualization, Bloomberg Terminal aesthetic. 16:9 aspect ratio, 2K resolution.

Climate Data Visualization

Create a clean, modern infographic titled "Global Temperature Rise 1880–2024" showing a simple area chart with a clear upward trend. Use a dark navy-to-midnight blue gradient background. The Y-axis shows temperature anomaly in °C (-0.5 to +1.5), the X-axis shows decades. Highlight 1980 and 2023 as key milestone markers with annotation callouts. Color the area chart in a warm gradient from pale yellow to deep orange-red as temperature increases. "Global Temperature Rise" in bold white Inter font at top. Source credit in small grey text at bottom. 16:9 aspect ratio, 2K resolution.

Key choices: area chart with color gradient creates visceral visual encoding of temperature increase; specific axis ranges give the model precise guidance; annotation callouts direct reader attention to meaningful inflection points.

Process / How-It-Works Diagram

A clean editorial infographic showing a 5-step supply chain process for a sustainable coffee brand. Steps: 1. Farm (Colombia), 2. Processing & Drying, 3. Quality Grading, 4. Roasting, 5. Packaging & Shipping. Each step shown as an icon in a rounded square, connected by horizontal arrows. Icons in warm terracotta color, arrows in pale gold, white background, body text in dark charcoal Inter font. Title: "From Farm to Cup" in bold. Minimal, editorial design reminiscent of a premium brand's sustainability report. 16:9 aspect ratio, 2K resolution.

Key choices: specific step names and geographic detail (Colombia) create authentic specificity; terracotta and gold palette communicates warmth and premium quality appropriate to the brand.

Comparison Table Infographic

Create a visual comparison infographic titled "Renewable Energy Sources — At a Glance." Compare three energy types: Solar, Wind, and Hydroelectric. For each, show: average cost per kWh, global capacity share (%), installation time, and geographic limitations. Use a three-column grid layout, one column per energy type. Color code: Solar in amber, Wind in sky blue, Hydroelectric in teal. Clean, modern icons for each category row. White background, Inter font, editorial data visualization style. 3:2 aspect ratio, 2K resolution.

Key choices: color-coding each column makes comparison scanning faster; specific data categories (cost, capacity, time, geography) prevent the model from improvising irrelevant metrics; white background maximizes readability for print or web use.

Timeline Infographic

A vertical timeline infographic titled "The History of Personal Computing" showing 8 key milestones from 1975 to 2025. Each entry has a year, a 2–3 word milestone title, and a small illustrative icon. Use alternating left-right text placement. Color scheme: deep purple gradient background with light lavender text and gold accent lines. Clean, editorial tech aesthetic. Icon style: simple line art in white. 2:3 portrait format, 2K resolution. Typography: bold Inter for years, regular Inter for descriptions.

Key choices: alternating left-right layout is a classic timeline design that creates visual rhythm; the vertical 2:3 format suits the sequential flow of a timeline; the purple-gold palette communicates technology and innovation without being generic.

14.5 Social Media

Instagram Story — Motivational

A bold, visually striking Instagram Story graphic in 9:16 format. Full-bleed background: a deep violet-to-midnight blue gradient. Centered white text in large, bold uppercase: "MAKE IT HAPPEN" on two lines, with the sub-text "Your effort compounds daily." below in thin, light grey. Subtle gold geometric lines radiate from behind the text. Clean, modern design aesthetic, minimal padding. 9:16 aspect ratio, 2K resolution.

Key choices: "full-bleed gradient" prevents edge cropping issues; uppercase bold for the hero message with thin lowercase for the sub-message creates natural typographic hierarchy; gold accents add premium quality without complexity.

LinkedIn Banner — Professional

A professional LinkedIn banner in 16:9 format. Clean architectural photography of a modern open-plan office space — warm natural light through floor-to-ceiling windows, clean white walls, minimal furniture. In the bottom-left third of the image, large white text reads "Strategy. Creativity. Growth." in a clean, bold sans-serif. A subtle dark gradient overlay on the left side ensures text legibility. Photography aesthetic: bright, airy, optimistic. 16:9 aspect ratio, 2K resolution.

Key choices: "dark gradient overlay on the left side" is specified deliberately to solve the common problem of white text on light backgrounds being unreadable; text in the lower-left leaves the right side clear for profile photo overlap.

Product Launch Announcement — Instagram Feed

A product launch social media graphic for Instagram feed in 1:1 format. White background. Centered product image: a slim, matte black smartwatch photographed from a slightly elevated front angle. Below the product, in clean black sans-serif: "INTRODUCING" in small caps, then "APEX WATCH" in large, bold display type, then "Available Now" in light, thin font with a subtle underline. Negative space is generous and intentional. Minimalist, Apple-product aesthetic. 1:1 aspect ratio, 2K resolution.

Key choices: white background and generous negative space let the product breathe — the product is the hero; the three-level typographic hierarchy (small caps label → bold product name → thin CTA) creates a classic announcement structure.

Twitter/X Header Banner

A graphic design for a Twitter/X header banner (3:1 very wide format). Abstract, painterly digital art: deep ocean waves in dark teal and navy, catching the light in silvery crests, stretching across the full width. The feeling is vast, quiet, powerful. No text. High-quality editorial illustration style, slightly desaturated, cinematic. 21:9 aspect ratio (use the widest available), 2K resolution.

Key choices: no text on a header banner prevents layout conflicts with profile photo and display name overlays; 21:9 is the closest available ratio to a Twitter header's actual ~3:1 crop; abstract art scales and crops well across devices.

14.6 Illustration

Children's Book — Full Spread

A full double-page spread illustration for a children's picture book. A small girl with bright red pigtails and yellow rain boots stands at the entrance of an enormous magical library, its shelves reaching impossibly high into the clouds above, illuminated by warm golden lantern light. Books float gently through the air. The girl looks up in wide-eyed wonder, her tiny figure emphasizing the magnificent scale of the space. Painterly, detailed illustration style with soft edges and a warm, luminous palette, reminiscent of Chris Van Allsburg's work. 16:9 aspect ratio, 2K resolution.

Key choices: "tiny figure emphasizing the magnificent scale" explicitly directs the compositional intent; floating books create the magical detail children's books require; Chris Van Allsburg reference is specific and well-defined (chiaroscuro, mysterious wonder).

Editorial Illustration — Opinion Piece

An editorial illustration for a technology opinion piece titled "The Algorithm's Grip." A human hand reaches up toward a glowing smartphone screen from below, fingers just touching the glass, while from inside the phone a metallic mechanical hand reaches down and gently wraps around the human wrist. Cool blue-grey tones with a single warm amber glow emanating from the phone screen. Graphic, slightly surreal style reminiscent of New Yorker editorial illustration — clean, considered, conceptually precise. 1:1 aspect ratio, 2K resolution.

Key choices: the reciprocal hands create the visual metaphor of mutual capture; warm amber from phone vs. cool surroundings emphasizes the screen as a separate, technological world; New Yorker editorial illustration style is a highly precise reference.

Fantasy Map Illustration

A hand-drawn fantasy cartography illustration of an island continent — an archipelago of interconnected islands with mountain ranges in the center, dense forests on the eastern coast, desert on the southern shore, and a network of rivers. Coastlines are rugged and detailed. Compass rose in the lower right. Decorative sea monsters in the open ocean. Location labels in elegant Old English calligraphy: "The Ashenwild," "The Silver Reaches," "The Ember Wastes." The overall aesthetic of a carefully illustrated antique map — sepia-toned parchment texture, fine ink linework, subtle watercolor fills. 3:2 aspect ratio, 2K resolution.

Key choices: specific geographic features (mountains center, forest east, desert south) create a coherent, real-feeling geography; parchment texture + sepia tone + ink linework together establish the antique map aesthetic precisely.

Technical Diagram Illustration — Cross-Section

A detailed, beautiful technical cross-section illustration of a beehive inside the hollow of a large oak tree, showing five labeled chambers: the entrance chamber, the food stores, the nursery cells, the queen's chamber, and the winter cluster. Illustrated in the style of a vintage Victorian natural history encyclopedia plate — fine pen-and-ink linework with soft, precise watercolor fills, cream/parchment background, elegant serif labels. Detailed, scientific, and visually beautiful simultaneously. 3:4 aspect ratio, 2K resolution.

Key choices: labeled chambers with specific names gives the model clear content to structure; Victorian natural history plate style is specific and consistent — fine ink + watercolor + cream background is a well-defined aesthetic the model renders accurately.

14.7 Architecture

Residential — Exterior Visualization

A photorealistic architectural visualization of a contemporary single-family home exterior. The house features a two-story mass in raw board-formed concrete with large floor-to-ceiling glazing, a cantilevered upper floor projecting over the entrance, and a green roof with low sedums. Set on a tree-lined suburban lot, wide lawn, late afternoon light creating long shadows across the board-form texture. Shot from a slight low angle, 24mm wide lens, f/8. Full-scene sharpness. 16:9 aspect ratio, 4K resolution.

Key choices: specific materiality (board-formed concrete, sedums on green roof) drives authentic architectural realism; late afternoon light with long shadows is the standard for architectural visualization — it emphasizes texture and form; 4K specified because architectural details at large print scale require it.

Interior Design — Living Space

A photorealistic interior architecture visualization of a modern, Japandi-inspired living room. White plaster walls, raw ash wood floors, a low platform sofa in undyed natural linen, a single large washi paper pendant lamp, a minimal stone coffee table. One entire wall is glass, opening to a bamboo-enclosed garden. Afternoon natural light filters diffusely through shoji screens on the left. Shot straight-on from seated eye level, 35mm, f/8. No clutter, no decorative objects — pure material and light. 3:2 aspect ratio, 4K resolution.

Key choices: "Japandi" is a precise design vocabulary reference; "no clutter, no decorative objects" is an explicit positive restraint that prevents the model from adding visual noise; seated eye level is the most natural viewing angle for a living room.

Urban Planning — Aerial Visualization

A photorealistic aerial architectural visualization of a proposed mixed-use urban development — a city block featuring a central public plaza with mature trees, surrounded by a mix of 6-story residential buildings and lower ground-floor retail frontages. People populate the plaza and sidewalks at human scale, giving sense of activity. Late morning sun from the southeast. Shot from a 45-degree aerial angle, looking northwest. Clean, contemporary architectural rendering aesthetic. 16:9 aspect ratio, 4K resolution.

Key choices: "people populate the plaza at human scale" is a crucial instruction for urban visualization — empty renderings feel dystopian; specific sun angle (southeast, late morning) gives the rendering a grounded temporal reality.

Historic Reconstruction

A photorealistic architectural visualization of the Roman Forum in Rome, reconstructed to its appearance circa 100 AD, at midday. The Via Sacra runs through the center. The Temple of Saturn is visible on the left, the Arch of Titus in the background, the Colosseum partially visible in the far distance. Populated with Roman citizens in period dress, merchants, and soldiers. Strong Italian midday light, white marble gleaming, terracotta and ochre tones in the surrounding buildings. Shot at ground level, moderate wide angle, 28mm equivalent. 16:9 aspect ratio, 4K resolution.

Key choices: specific date (100 AD) anchors the reconstruction; named structures (Temple of Saturn, Arch of Titus) provide verifiable architectural anchors; populated scene with period-accurate people turns a technical render into a living image.

14.8 Food Photography

Fine Dining — Plated Dish

A high-end fine dining food photograph of a plated dish: a single perfectly seared scallop, golden-brown crust, placed on a swoosh of pale celery root purée on a matte white ceramic plate. Three tiny drops of bright green herb oil, one pickled micro-shallot, a single chervil sprig. Photographed from directly above (bird's-eye view), single soft light source from the upper left creating subtle shadows. The focus is total — every element is razor-sharp (f/8). Clean white background. Shot on Hasselblad, 100mm macro lens. 1:1 aspect ratio, 4K resolution.

Key choices: bird's-eye view is the canonical angle for this kind of architectural fine dining photography; "every element is razor-sharp" is an explicit instruction that prevents the default shallow-DOF look; naming micro-shallots and chervil gives the model specific, real garnish vocabulary.

Comfort Food — Editorial Lifestyle

A warm, inviting lifestyle food photograph of a cast iron skillet of bubbling shakshuka — vivid tomato sauce, poached eggs with set whites and runny yolks, topped with crumbled feta, fresh parsley, and red chilli flakes. The skillet sits on a worn wooden kitchen table with a linen cloth napkin folded beside it, and a small glass of orange juice partially in frame. Photographed from above at a slight angle — 3/4 overhead view. Soft natural window light from the left. Warm, cozy tones. Editorial food photography aesthetic, slightly casual and lived-in. 3:2 aspect ratio, 2K resolution.

Key choices: "runny yolks" is a specific textural note — it communicates freshness and triggers appetite; the "3/4 overhead view" avoids the perfectly top-down look, creating a more natural human perspective; "worn wooden table" and "linen napkin" add domestic warmth.

Baking — Product Close-Up

A close-up food photograph of a freshly cut slice of artisan sourdough bread, revealing an open, irregular crumb structure. The slice rests on a dark slate surface. Morning natural light from a side window creates long, dramatic raking shadows that emphasize the crumb's texture. A smear of grass-fed butter is melting visibly on the cut face. Shot on a 100mm macro lens, f/4. Fujifilm Velvia-inspired color grading emphasizes the gold of the crust and the ivory of the crumb. 3:2 aspect ratio, 2K resolution.

Key choices: "raking shadows" (side light creating shadows that emphasize texture) is the correct technique for demonstrating bread crumb — it's a deliberate, professional lighting choice; Velvia color reference enriches the golden crust tones.

Beverage — Cocktail Photography

A professional bar cocktail photograph of a Negroni served in a large round crystal rocks glass with a single oversized ice cube. The deep amber-red of the drink is backlit by a warm amber point light source behind, creating a luminous internal glow. A neatly curled orange twist rests on the rim, catching specular highlights. The glass sits on a dark slate bar surface. Additional out-of-focus bar elements are visible in the background bokeh. Shot on 85mm, f/1.8, shallow depth of field. 1:1 aspect ratio, 2K resolution.

Key choices: backlighting a cocktail is the professional technique for making liquid color pop — specified explicitly; "large round crystal rocks glass" is specific to the Negroni aesthetic; f/1.8 creates the soft background bokeh that contextualizes without competing.

15. Common Mistakes → Better Versions

Learning prompt writing is partly learning to recognize weak patterns and replace them. Here are eight before/after comparisons that illustrate the most common failure modes — and how to fix them.

Mistake 1 — Keyword Soup

Why the weak version fails: Tags like "masterpiece" and "ultra detailed" are legacy Stable Diffusion hacks that mean nothing to a reasoning model. Without narrative structure, the model has no spatial relationships, mood intent, or compositional direction. The strong version gives it everything.

Mistake 2 — Negative Framing

Why the weak version fails: The model processes concepts it's told to avoid — generating them and then attempting to suppress them produces artifacts and inconsistencies. Positive descriptions of the desired state are unambiguous.

Mistake 3 — Vague Subject

Why the weak version fails: "A chef cooking" could be anyone, doing anything. The strong version specifies age, appearance, specific action, specific food, specific environment, and emotional quality — creating a real, specific human moment.

Mistake 4 — No Camera Direction

Why the weak version fails: "Dramatic" is subjective and unspecified. The strong version defines what makes it dramatic — the foreground rocks, the low angle, the golden peak in the distance. Camera direction is compositional decision-making, not optional detail.

Mistake 5 — Not Quoting Text

Why the weak version fails: Without quotes around text content, the model may render approximate or incorrect text. The strong version also specifies typography for each text element — font weight, color, and style — creating a legible, designed poster.

Mistake 6 — Contradictory Instructions

Why the weak version fails: "Photorealistic" and "Studio Ghibli animation" are contradictory rendering goals — the model will average between them and satisfy neither. Choose one visual register and commit to it.

Mistake 7 — Re-Rolling Instead of Editing

Why the weak approach fails: Starting over from scratch discards all the good work the model has already done. A targeted edit instruction preserves what's working while precisely fixing what isn't — and in a thinking model with session context, this is dramatically more efficient.

Mistake 8 — Ignoring Reference Image Roles

Why the weak version fails: Without role assignments, the model guesses what to do with each image. It may blend features inappropriately — using the product's lighting as the scene lighting, or borrowing shapes from the background into the product. Explicit roles give unambiguous direction.

16. Prompt Building Workshop

One of the most valuable skills in prompt writing is the ability to start from a vague idea and systematically refine it into a production-ready prompt. Here's a complete workshop that walks through five iterations, showing the thinking at each step.

Starting Idea: "I want an image of a woman reading a book"

That's a valid creative starting point. Now let's build it into something real.

Iteration 1 — Apply the Basic Framework

Apply the 5-part Text-to-Image formula: Subject + Action + Location + Composition + Style.

A woman reading a book in a café. Medium shot. Photorealistic.

What we have: A technically valid prompt. The model will produce something. But it will be generic — average lighting, average setting, average styling.

What's missing: Who is this woman? What kind of café? What's the emotional quality? Why are we looking at this image?

Iteration 2 — Specify the Subject

Make the subject a real, specific person with visual identity.

A woman in her late 30s with loose, dark wavy hair and a relaxed, intellectual bearing, wearing a soft sage green linen shirt, reading a worn paperback novel in a café. Medium shot. Photorealistic.

What we've added: Age, hair description, clothing with color and material, emotional quality ("relaxed, intellectual"), and the book is now a "worn paperback novel" — which implies something about character.

What's still missing: Where exactly is she? What's the light? What's the composition?

Iteration 3 — Build Out the Environment and Lighting

Make the setting real and specific, and make a deliberate lighting decision.

A woman in her late 30s with loose, dark wavy hair and a relaxed, intellectual bearing, wearing a soft sage green linen shirt, reading a worn paperback novel. She sits at a small round marble table near a large window in a traditional Parisian café — mirrors on the walls, bentwood chairs, afternoon light filtering through gauze curtains. The light is soft, directional from the window, casting a gentle glow on her face. Medium shot. Photorealistic.

What we've added: Specific Parisian café setting with real architectural details; afternoon window light with specific quality; the gauze curtains soften the light in a way that establishes a mood.

What's still missing: Camera direction and style reference.

Iteration 4 — Add Camera Control and Style

This is where the image stops looking like a snapshot and starts looking like a photograph.

A woman in her late 30s with loose, dark wavy hair and a relaxed, intellectual bearing, wearing a soft sage green linen shirt, reads a worn paperback novel at a small round marble table near a large window in a traditional Parisian café — mirrors on the walls, bentwood chairs, afternoon light filtering through gauze curtains. The light is soft and directional from the window, casting a gentle glow on her face and the pages of the book. Shot at medium distance, slightly above eye level, 50mm lens, f/2.0 — the background café is pleasantly out of focus. Fujifilm Pro 400H color grading — soft, slightly desaturated, warm skin tones. 3:4 portrait aspect ratio, 2K resolution.

What we've added: Specific lens (50mm), aperture (f/2.0), camera angle (slightly above eye level), explicit bokeh instruction for the background, and a precise color grading reference (Fujifilm Pro 400H — which gives exactly the soft, slightly faded look appropriate for a Parisian afternoon).

What's still missing: A sense of purpose, and we can sharpen one more element.

Iteration 5 — Add Purpose and Final Polish

The "why" — what is this image for? — focuses the final creative decisions.

A quiet, literary lifestyle photograph. A woman in her late 30s with loose, dark wavy hair and a relaxed, intellectual bearing, wearing a soft sage green linen shirt, is absorbed in reading a worn paperback novel — you can see she's nearly at the end. She sits at a small round marble table near a large window in a traditional Parisian café: worn bentwood chairs, mirrors with gold frames, the sound of quiet conversation implied by the scene. Afternoon light filters through gauze curtains, casting soft, directional window light across her face and the pages. Shot at medium distance, slightly above eye level, 50mm lens, f/2.0, background café pleasantly defocused. Fujifilm Pro 400H color grading — soft, desaturated, warm. For use in an editorial lifestyle campaign for a literary magazine. 3:4 portrait aspect ratio, 2K resolution.

Final additions: "you can see she's nearly at the end" — a tiny, specific narrative detail that creates genuine character; "the sound of quiet conversation implied" — a synaesthetic note that tells the model to make the background feel occupied without being distracting; "for use in an editorial lifestyle campaign for a literary magazine" — purpose context that orients every compositional trade-off.

The five iterations compared:

The lesson: A production-quality prompt is typically 100–175 words. You don't need to start there — you build there. Each iteration adds a layer of intent that makes the final image more specific, more emotional, and more useful.

17. Cheat Sheet — Quick Reference

Cut this out, print it, pin it up. Everything you need on one page.

The Core Formula

[Subject + specific visual identity]
[Action or pose]
[Location/environment with details]
[Lighting specification]
[Camera: shot type + lens + aperture]
[Color grading / film stock / style reference]
[Output: aspect ratio + resolution]

Non-Negotiable Rules

Aspect Ratio Quick Reference

Resolution Decision

Lighting Quick Reference

Film Stock / Color Reference

Focal Length Reference

Reference Image Roles (Copy-Paste)

Use Image A for [the character's visual identity / facial features].
Use Image B for [the color palette and mood].
Use Image C for [the background environment].
Use Image D for [the product shape and proportions].
Use Image E for [the art style and rendering treatment].
Keep [the person's face / the product design / the color palette] identical to the reference.

The Grid Trick

Request a 2×2 grid in a single 2K generation = 4 distinct sub-images at ~1MP each. Best for rapid variation exploration.

Generate a 2×2 grid of four variations of [concept], each with a different [variable].

When to Use Pro vs. Nano Banana 2

Prompt Length Guide

Guide compiled and expanded from: Google Cloud Blog — Ultimate Prompting Guide for Nano Banana, Google DeepMind Gemini Image Prompt Guide, Google Blog — Prompting Tips for Nano Banana Pro, DEV Community — Nano Banana Pro Prompting Guide: Strategies, Max Woolf's Blog — Nano Banana Pro Deep Dive. Last updated: March 2026.