12 Nano Banana Prompts That Generate On-Brand Product Images First Try

Most operators using Nano Banana (the operator nickname for Google's Imagen 3 on Vertex AI) burn 20-30 minutes per usable image because they're writing prompts from scratch every time. The output drifts. The hit rate sits around 25-30%. Three to four re-rolls per usable image. Not viable for production volume.

The fix is a small library of paste-and-ship prompt recipes — positive-prompt skeleton + paired negativePrompt — for the 12 most common operator use cases. Drop in your subject, generate, ship. Hit rate jumps to ~80% first try.

This article is the 12 recipes. Plus the structure that makes them work.

How To Use Each Recipe

Every recipe has two parts:

Positive prompt — describes what you want.

negativePrompt — describes what to exclude.

Both go in the Vertex AI API call:

`` { "prompt": "[positive prompt with your subject inserted]", "negativePrompt": "[matching negativePrompt for the use case]", "sampleCount": 1 } ``

The negativePrompt parameter is the unlock. Most prompt guides skip it. Without it, Nano Banana defaults to messy outputs (watermarks, distorted hands, multiple subjects, text overlays). With the right negativePrompt, those failure modes vanish.

See the negativePrompt deep-dive for why this matters.

Recipe 1 — Product Mockup (Single Product)

Positive:

"A [product name] photographed on a [color] seamless background, centered composition, soft directional lighting from upper left, sharp focus, magazine quality, high resolution, photorealistic, commercial product photography style"

negativePrompt:

"text, watermark, blurry, distorted, low quality, multiple products, additional items in frame, people, hands, busy background"

Recipe 2 — Hero Image (Marketing / Landing Page)

Positive:

"A [subject description with action and emotion], natural soft light, contemporary editorial photography style, shallow depth of field, high resolution, photorealistic, magazine-quality composition, warm color grading"

negativePrompt:

"text, watermark, blurry, distorted, low quality, stock photo aesthetic, generic, cluttered, busy, dated style, low resolution, multiple subjects"

Recipe 3 — Portrait (Single Person, Professional)

Positive:

"A [age range] [gender] with [hair description] and [distinctive feature], wearing [attire], professional headshot, neutral [color] background, soft window light from the side, sharp focus on the eyes, contemporary editorial photography, high resolution"

negativePrompt:

"text, watermark, blurry, distorted, low quality, extra fingers, deformed hands, multiple people, group photo, cartoon, illustration, painting, low resolution"

Recipe 4 — Lifestyle Scene (Person Using Product)

Positive:

"A [person description] [doing specific action with product] in a [setting description], natural lifestyle photography, warm afternoon light, candid composition, high resolution, magazine-quality, shallow depth of field, photorealistic"

negativePrompt:

"text, watermark, blurry, distorted, low quality, extra fingers, deformed hands, multiple people unless requested, posed stiffly, awkward composition, low resolution"

Recipe 5 — Social Media Graphic (Hook-Style)

Positive:

"A clean conceptual composition showing [concept or metaphor], minimalist design, single focal point, bold color contrast between [color A] and [color B], modern editorial graphic style, high resolution, clean negative space"

negativePrompt:

"text overlay, watermark, blurry, distorted, low quality, busy background, multiple competing focal points, dated aesthetic, cluttered, generic"

Recipe 6 — Food Photography

Positive:

"A [dish name] photographed on [plate/board description], overhead 45-degree angle, natural window light, warm color tones, minimal styling, high resolution, magazine-quality food photography, shallow depth of field"

negativePrompt:

"text, watermark, blurry, distorted, low quality, people, hands, busy background, dirty plates, garnish overload, dated photography style, plastic-looking food"

Recipe 7 — Landscape / Environment

Positive:

"[Landscape description] during [time of day], natural lighting, clean composition, high resolution, photorealistic, contemporary editorial style, atmospheric depth, single focal point"

negativePrompt:

"text, watermark, blurry, distorted, low quality, people, vehicles, signs, clutter, garbage, modern logos, brand names, power lines"

Recipe 8 — Interior / Architecture

Positive:

"[Room type] interior in [design style], natural light from large windows, clean composition, contemporary photography style, high resolution, photorealistic, magazine-quality architectural photography, warm color grading"

negativePrompt:

"text, watermark, blurry, distorted, low quality, people, clutter, dated furniture, busy patterns, low ceilings unless requested, harsh lighting"

Recipe 9 — Editorial Illustration

Positive:

"Editorial illustration showing [concept], minimalist style, limited [color] palette, clean geometric composition, contemporary publication style, single focal point, conceptual metaphor visualized clearly"

negativePrompt:

"text, watermark, blurry, distorted, low quality, photorealistic, busy background, multiple competing elements, dated illustration style, generic stock illustration"

Recipe 10 — Infographic / Data Style

Positive:

"Modern infographic-style illustration representing [concept], clean geometric shapes, limited color palette of [color A], [color B], and white, abstract data visualization aesthetic, contemporary editorial style"

negativePrompt:

"text, watermark, blurry, distorted, low quality, photorealistic, busy chart, multiple competing data points, dated infographic style, generic stock"

Recipe 11 — Pattern Background

Positive:

"Seamless repeating pattern of [motif description] in [color palette], minimalist composition, modern textile design style, even distribution, high resolution, suitable for background or product surface"

negativePrompt:

"text, watermark, blurry, distorted, low quality, photorealistic, faces, recognizable objects, asymmetric composition, harsh transitions"

Recipe 12 — Explainer Scene (Conceptual)

Positive:

"A conceptual scene showing [process or relationship], clean composition, modern illustration style, limited [color] palette, single clear visual metaphor, contemporary editorial style, suitable for blog hero image"

negativePrompt:

"text, watermark, blurry, distorted, low quality, busy composition, multiple competing concepts, generic clip art, dated illustration style"

How To Customize Per Brand

The recipes above are universal. To brand-customize without breaking the hit rate, layer 2-4 brand-specific tokens at the START of the positive prompt:

Example for a brand with sage-green palette and warm editorial photography style:

"Sage green color palette, warm editorial photography style, soft natural light. A [product name] photographed on a cream seamless background, centered composition, soft directional lighting from upper left, sharp focus, magazine quality, high resolution, photorealistic, commercial product photography style"

The brand tokens land in the highest-weighted position of the prompt. The recipe structure stays intact. Hit rate holds.

The Cost Math

Per recipe with 80% first-try hit rate:

• Nominal cost per generation: $0.04
• Effective cost per usable image: $0.04 / 0.80 = $0.05

For a typical operator project (50 images for a brand refresh): ~$2.50 in API spend.

For an ongoing client commitment (30 images/month at $299/month — see first AI image client article): ~$1.50/month in API spend on a $299 deliverable. ~99.5% gross margin.

When Recipes Fall Short

A few situations where you'll need to deviate:

Highly specific brand composition. "Always shoot at 35mm equivalent, always center-weighted, always cool color cast" — these specifics need 4-6 additional brand tokens beyond the recipe.

Character consistency across images. Use the 6-trait character lock pattern from Veo. Same approach works in Nano Banana.

Specific brand objects (your specific bottle, your specific packaging). Use image conditioning (Imagen 3 supports a reference image input). Pure text-to-image won't produce your specific SKU.

Multi-subject composed scenes. The recipes are tuned for single-subject. Multi-subject needs more detailed positional prompting.

The Cross-Sell

The Nano Banana (Imagen 3) for Operators guide ($5.99) includes all 12 prompt recipes as paste-and-ship files, the negativePrompt library, the per-niche cost calculator, brand-customization templates, and the client offer template covered in the first-client article.

$5.99 once. Most operators recoup the cost on the first generation session where the recipes save 60+ minutes of prompt-tuning.

The actionable next step: pick the recipe matching your most-common use case. Drop in your subject. Hit generate. Notice the hit rate. The improvement is usually obvious by image 2.