Why can other people turn their Nano Banana Pro images straight into posters, while yours still feel a little off? The model is rarely the problem. Almost always, it comes down to the prompt.
This guide gives you three things: a reusable prompt formula, 10 ready-to-copy examples organized by use case, and the most common beginner mistakes to avoid. For more inspiration, explore our prompt library, featuring 2000+ prompts. You can also paste the examples into our online AI image generator and refine them as you read.
1. Nano Banana, Nano Banana 2, and Nano Banana Pro: How Should Your Prompts Differ?
First, a quick clarification: these models belong to the same image model family, but they were released at different times and serve different purposes. Their names do not represent three generations in numerical order.
- Nano Banana: Released in August 2025, its official model name is Gemini 2.5 Flash Image. It is an earlier model in the family, suited to everyday image generation and editing.
- Nano Banana 2: Released in February 2026, its official model name is Gemini 3.1 Flash Image. It combines several Pro capabilities with Flash speed. At launch, Google announced its rollout across multiple modes in the Gemini app. Check the current product interface for available models.
- Nano Banana Pro: Released in November 2025, its official model name is Gemini 3 Pro Image. It prioritizes image quality and factual accuracy, making it suitable for professional work that demands more detail.
These differences help determine how simple or detailed your prompts should be:
| Comparison | Nano Banana | Nano Banana 2 | Nano Banana Pro |
|---|---|---|---|
| Positioning | Earlier model for everyday generation and editing | Balance of speed and quality for rapid iteration | High-fidelity creation for professional refinement |
| Prompt style | Start with 1–2 clear, concise sentences | Add task-specific details while keeping iteration efficient | For complex tasks, describe the camera, lighting, materials, and layout in detail |
| Text rendering | Start by testing short text | Supports clearer text rendering and multilingual content | Strong at complex text layouts, with multilingual support and font-style descriptions |
| Reference images / consistency | Supports reference-based editing | Supports multiple reference images and subject consistency | Official capabilities include blending up to 14 reference images and maintaining consistency for up to 5 people |
| Resolution | Approximately 1K output; dimensions vary by aspect ratio | Supports 512px, 1K, 2K, and 4K output | Supports 1K, 2K, and 4K output |
| Web search | Not supported | Supported | Supported |
| Generation speed | Suitable for routine generation and editing | Designed for fast generation and iteration | Suited to tasks that prioritize detail and precision |
The simple takeaway:
- Nano Banana: Keep prompts short and precise. Its capabilities are more basic, so one or two sentences describing the subject and style are enough for lightweight generation and simple edits.
- Nano Banana 2: Use a moderate level of detail. Clearly state the key requirements for everyday images, batches of drafts, and other efficiency-focused tasks.
- Nano Banana Pro: Be more detailed. Describe the camera, lighting, materials, and composition for precision-focused work such as posters, infographics, and layouts with longer text.
One more option: Google later introduced the lighter Nano Banana 2 Lite, focused on maximum speed and low cost. It can be a starting point for quick drafts when fine detail is not a priority.
2. The Core Formula for Better Prompts
You do not need to memorize complicated theory. One formula covers most everyday use cases:
Subject + scene/environment + composition and camera + lighting + style + finishing details
Here is the difference in practice:
- ❌ Vague:
A young woman in a café - ✅ Using the formula:
A woman in her 20s reading by the window of a vintage café, 35mm film look, afternoon sunlight coming in from the side, shallow depth of field with a blurred background, cinematic warm tones
The second prompt does not simply add more adjectives. It specifies camera language, lighting direction, and color palette—key variables that shape the final image.
A few practical principles:
- Change one variable at a time: When the result misses the mark, first check whether the subject is unclear or the style instructions conflict. Do not change ten things at once.
- Choose concrete terms over abstract ones: Instead of “beautiful lighting,” try “golden hour side-backlighting” or “warm dusk light.”
- Use reference images: Especially with Nano Banana Pro, uploading a reference image communicates a style more precisely than words alone.
- Write in your preferred language and keep specialist terms in English: Whether you write in Chinese, Japanese, Spanish, or another language, English photography terms such as “bokeh” and “top-down” are easier for the model to recognize.
To avoid rebuilding your prompt from scratch every time, use our regularly updated prompt library. Prompts are organized by use case and ready to copy and paste.
3. 10 Prompt Examples You Can Adapt
Here is one representative prompt for each of 10 use cases. Copy and customize them, or write your own using the same “subject + scene + lighting + style” structure.
- Portrait photography: A female architect around 30 years old, wearing an ivory suit, standing in front of a modern glass building, waist-up composition, 85mm lens, evening side-backlighting, blurred background, cinematic look, warm gray tones.
- E-commerce product photography: A pair of silver wireless earbuds centered on a light gray background, shown at 45°, soft overhead lighting, delicate metallic highlights, minimalist premium e-commerce style, clean composition.
- Infographic/poster: An AI product launch poster with a laptop in the center, headline “Create Without Limits,” subtitle “AI Creation, Reimagined,” blue tech-inspired style, plenty of white space, 16:9.
- Illustration: A small red vintage train traveling through an autumn valley, misty mountains in the distance, soft morning light, watercolor illustration, visible paper texture, orange and green palette, storybook atmosphere.
- Image editing: Replace the original background with a Tokyo street at night after rain. Keep the person’s pose and appearance unchanged. Let neon light fall naturally on the person, with reflections on wet pavement and a softly blurred background.
- Character consistency: Use the person in the reference image to create a scene of her reading beside a train window. Keep her facial features and hairstyle consistent, 50mm lens, natural afternoon light, realistic photography.
- Architecture: A modern minimalist home built with concrete and natural wood, large floor-to-ceiling windows, a valley visible in the distance, wide-angle composition, soft early-morning light, clean and refined space.
- Food photography: A freshly baked French croissant on a ceramic plate, with a hot latte and blueberries nearby, 45° overhead angle, morning window light, warm tones, shallow depth of field, refined food photography.
- Social media cover: A YouTube thumbnail about “I Tested 5 AI Image Models,” a surprised person on the left, 5 image thumbnails on the right, blue and yellow palette, strong tech feel, 16:9.
- Logo design: Design a logo for “Lumio” that combines the letter L with a beam of light, simple geometric lines, modern SaaS style, primarily black and white with light blue accents, suitable for an icon and brand presentation.
These 10 examples are a starting point for common tasks, including portraits, products, posters, illustrations, editing, and character consistency. For a more complete collection, visit our regularly updated prompt library, featuring over 2000 ready-to-copy prompts organized by category. Bookmark it and explore at your own pace.
4. Advanced Techniques for More Precise Prompts
1. Learn a few photography terms
Wide-angle Macro Shallow depth of field/bokeh Backlighting Overhead lighting Rule of thirds Symmetrical composition Film grain Long exposure — Adding these terms helps the model understand the visual result you have in mind.
2. Control text rendering precisely
When generating text with Nano Banana Pro, specify the exact wording in quotation marks, the font style, such as “handwritten” or “bold sans serif,” and the position, such as “centered” or “top left.” Specific instructions reduce ambiguity, but you should still check every character afterward.
3. Make use of web search
Nano Banana Pro and Nano Banana 2 can use search results when search grounding is supported and enabled. Examples include “Create an information card showing today’s weather in Tokyo” or “Create an illustrated step-by-step tiramisu recipe.” Whether a search actually runs depends on the platform and its settings. Adding “today” to a prompt does not mean the model has checked current information. On our site, you can first add verified data to your prompt, then ask the model to arrange it into an image.
4. Not sure how to write a prompt? Start with a reference image
When you find an image you like but cannot put its style into words, use our image-to-prompt tool. Upload the image to generate a description, then adapt it to your needs instead of starting from scratch.
5. Be specific about character consistency
Use clear reference images from different angles, such as front and side views. Explicitly ask to “keep facial features and clothing details consistent.” This communicates your requirements more clearly than simply writing “the same person.”
5. Common Mistakes to Avoid
- Overly vague prompts: “A beautiful image” gives the model little direction. Include at least a subject, a scene, and a style.
- Too many unrelated elements: The more you add, the easier it is to lose the main focus. Break complex requests into stages and refine the image through successive edits.
- Ignoring the aspect ratio: Posters, avatars, and banners need different proportions. Check the dimensions before generating.
- Garbled text: Try Nano Banana Pro for text-heavy images, put the exact wording in quotation marks, and check every character afterward.
- Inconsistent terminology: Use a consistent term for each concept, such as “close-up,” rather than mixing conflicting camera or style instructions in one prompt.
Conclusion
Good prompts are not a matter of guesswork. They come from practicing the “subject + scene + camera + lighting + style” formula. These 10 examples are just the beginning. Explore the full collection of 2000+ prompts in our regularly updated prompt library, or paste a prompt into our online AI image generator to test it right away without switching platforms.
