Nano Banana (Google's image model) is the best price-per-quality image generator of 2026 by a margin. But like every AI model it's sensitive to prompt structure. These are the 17 patterns that consistently produce publishable output at our shop.

The structure that always works

Most successful Nano Banana prompts follow this shape:

[SUBJECT] + [ACTION/POSE] + [ENVIRONMENT] + [LIGHTING] + [CAMERA/LENS] + [STYLE]

Example:

"A red fox (SUBJECT) sitting upright (ACTION) in fresh snow at the edge of a pine forest (ENV), backlit by golden hour sunset (LIGHT), shot on 85mm with shallow depth of field (CAMERA), photo-realistic (STYLE)"

17 patterns, copy-paste ready

Product photography

  • "[Product] on [surface], studio softbox, clean white background, advertising photography, macro lens"
  • "[Product] floating in mid-air, dramatic rim lighting, e-commerce hero shot, ultra-sharp"

Portrait

  • "[Subject description], natural daylight from window, shallow depth of field, 85mm lens, editorial portrait"
  • "[Subject], moody cinematic lighting, color grade teal and orange, shot on Arri Alexa"

Landscape / nature

  • "[Location] at golden hour, wide angle, volumetric light rays through [feature], National Geographic style"
  • "[Scene] in mist, low-key lighting, minimal composition, fine art photography"

Editorial illustration

  • "Flat vector illustration of [scene], soft pastel palette, rounded shapes, SaaS landing page hero"
  • "Minimal line drawing of [subject], single color, editorial newspaper illustration"

Concept art

  • "[Setting], cinematic concept art, wide establishing shot, dramatic scale, matte painting"
  • "[Character] in [environment], key frame concept art, Feng Zhu style"

Food

  • "[Dish] on [plate], overhead flat lay, natural light from left, food photography, 45mm lens"
  • "[Dish] being [action — e.g. sliced], freeze-frame macro, rich color, editorial food shot"

Architecture

  • "[Building style] in [location], blue hour, long exposure, wide angle, architecture photography"
  • "Interior of [space], natural light, warm tones, interior design magazine cover"

Abstract / texture

  • "Macro close-up of [material/surface], natural light, ultra-sharp, abstract texture"
  • "Liquid [color] morphing into [shape], studio lighting, product render"

Event / lifestyle

  • "[Activity] happening in [setting], candid moment, film grain, lifestyle photography"

Three mistakes that waste credits

1. Over-specifying

More than ~60 words of prompt starts to confuse the model. Trim adjectives. Pick two lighting terms, not five.

2. Asking for text

Nano Banana is weak on embedded text (menus, signs, labels). If your image needs text — use Flux Kontext Pro instead. We route automatically based on prompt content in our own product.

3. Negatives in the prompt body

"A beach but no people" often includes people anyway. Instead: just describe what you want. "Empty tropical beach at sunrise" works; "Beach without people" doesn't.

Quick reference card

Screenshot, save, reuse:

  • 📸 Realism: add "photo-realistic, 35mm film, Kodak Portra" at the end
  • 🎨 Illustration: specify medium — "flat vector, watercolor, line drawing, oil painting"
  • 💡 Lighting: "golden hour", "blue hour", "studio softbox", "natural window light"
  • 📷 Lens: "24mm wide", "50mm standard", "85mm portrait", "macro" — gives realistic optics
  • ✂️ Composition: "rule of thirds", "centered", "leading lines", "flat lay overhead"

Copy the pattern list into your team's internal doc and you'll stop burning credits on re-rolls.