Nano Banana 2 Prompts: Copy Templates for AI Images
Nano Banana 2 prompts should be specific, visual, and easy to iterate. The goal is not to write the longest prompt. The goal is to tell the model what to create, what to preserve, and what to avoid.
Prompt library
Copy a Nano Banana 2 prompt and generate it in BananaGen
Pick one template below, replace the bracketed details, then open BananaGen to create a real image.
Use these prompts in BananaGenNano Banana 2 Prompt Format
Every prompt below is built from the same six slots. Fill the ones that matter and drop the rest — an empty slot is better than a vague one.
| Slot | What it answers | Example |
|---|---|---|
| Subject | What is in the frame | a matte black espresso machine |
| Scene | Where it sits | on a walnut counter, morning kitchen |
| Lighting | How it is lit | soft window light from the left |
| Camera | Framing and lens feel | eye-level, 50mm, shallow depth of field |
| Style | The visual register | editorial product photography |
| Constraints | Keep and avoid | keep the label readable, no text, no watermark |
Written out, that is the reusable builder:
Create [subject] in [scene]. Use [camera/composition]. Lighting should be [lighting]. Style should be [style]. Keep [constraints]. Avoid [things to avoid].
Save that one line. Everything else on this page is it filled in differently.
A filled-in example:
A premium matcha drink can on a glossy yellow studio surface, soft shadow, macro product photography, front-facing composition, clean label area, no extra text, no distorted logo.
Put Yourself Into a Scene (Identity-Preserving Edit)
The most-requested edit is dropping a real person or product from an uploaded photo into a new setting without the face or label changing. State the identity lock explicitly — the model will drift otherwise.
Use the uploaded photo as the exact identity and subject reference. Place the subject naturally in the [position] of a [aspect ratio] [scene]. Preserve the face, hairstyle, clothing, pose, body proportions, and expression exactly. Replace the background with [new background], with [lighting]. Maintain realistic perspective, shadows, and depth, and blend the edges seamlessly. Do not alter the subject.
Best for: travel composites, headshots against a new backdrop, product staging, and profile images.
Product Image Prompt
A luxury [product] on a warm stone surface, soft golden rim light, realistic reflection, clean ecommerce composition, high-end product photography, sharp product edges, no extra objects, no watermark.
Best for: ecommerce images, Amazon-style product shots, ad creatives, and landing page visuals.
Portrait Prompt
A professional editorial portrait of [person description], natural skin texture, warm studio light, 85mm lens look, calm confident expression, clean background, preserve facial identity, no distorted hands, no watermark.
Best for: creator profiles, avatars, team portraits, and campaign concepts.
Image-to-Image Edit Prompt
Keep the main subject, pose, and lighting unchanged. Replace the background with [new background], make the scene realistic, preserve edges and shadows, no extra objects, no text, no watermark.
Best for: product background swaps, profile edits, furniture staging, and social visuals.
Ad Creative Prompt
A high-converting ad visual for [product/service], bold composition, warm yellow accent color, premium studio lighting, realistic product focus, clean space for headline, modern ecommerce campaign style, no readable text.
Best for: paid ads, launch banners, email hero images, and A/B test creatives.
Social Media Prompt
A vertical 9:16 social media visual showing [subject], bright clean background, cinematic lighting, creator-friendly composition, polished but natural style, strong focal point, no text, no watermark.
Best for: TikTok covers, Instagram posts, Reels thumbnails, and creator campaigns.
Logo Concept Prompt
A clean banana-inspired AI image generator logo concept, modern minimal mark, soft yellow and black palette, simple vector style, high contrast, no small unreadable text, centered on plain background.
Best for: early brand exploration. Use a designer for final trademark-safe logo work.
Video Thumbnail Prompt
A bold YouTube thumbnail about Nano Banana 2, bright yellow composition, expressive AI image examples, large empty space for title text, high contrast, modern creator style, dramatic lighting, no readable text.
Best for: thumbnail concepts and layout exploration before final text design.
Common Prompt Mistakes
Stacking styles. "Cinematic, anime, watercolor, photorealistic, 3D render, oil painting" gives the model six conflicting instructions and it will average them into mush. Pick one.
Asking for small text. Labels, logos and poster copy rarely survive generation cleanly. Leave empty space in the composition and add typography afterwards.
Editing with "make it better". In image-to-image the model needs both the change and the boundary. Replace the background with a clean modern kitchen, keep the product shape, label, and camera angle unchanged works; make it better does not.
Changing three things at once. When a draft is close, change one variable per generation. Otherwise you cannot tell which edit helped.
Prompt Improvement Checklist
- Name the exact visual job: product, portrait, poster, ad, logo, thumbnail, or scene.
- Describe lighting because it strongly changes perceived quality.
- For edits, state what must stay unchanged.
- Choose one visual style instead of mixing many styles.
- Add short negative constraints such as “no text” or “no watermark”.
- Generate a few drafts, then refine the strongest result.
Turn a template into an image
The fastest way to learn prompting is to run one template, inspect the result, then change only one variable at a time. New accounts get 12 free credits, which covers two Nano Banana 2 images - enough to run a template and one revision.
Open BananaGen and generateRelated pages: Nano Banana 2 vs Nano Banana Pro, How to use Nano Banana 2, Nano Banana 2 free, and Nano Banana 2 app.