Ultimate Prompt Guide
Everything you need to know about creating stunning AI images with Gemini
Based on Google's official documentation, this guide will help you master image generation and editing with Nano Banana (Gemini 2.5 Flash) and Nano Banana Pro (Gemini 3 Pro Preview).
The Golden Rule
Here's the secret to great AI images: **describe the scene, don't just list keywords**. The model's real strength is its deep language understanding. A narrative paragraph almost always creates better, more coherent images than a string of random words.
mountain, sunset, clouds, dramaticA majestic mountain peak bathed in golden sunset light, with dramatic clouds rolling across the skyProven Strategies for Image Generation
These techniques come straight from Google's guide and really work
1. Realistic Scenes
Want photorealistic results? Think like a photographer. Mention camera angles, lens types, lighting, and details to guide the model.
Tips:
- •Specify camera angles (wide-angle, close-up, bird's eye view)
- •Name lens types (35mm, macro, telephoto)
- •Describe lighting (golden hour, soft diffused light, dramatic shadows)
- •Add details (shallow depth of field, bokeh, film grain)
Example Prompt:
A close-up portrait of an elderly Japanese potter at work, photographed with a 35mm lens at f/1.8. Soft natural window light from the left creates gentle shadows. Shallow depth of field with beautiful bokeh in the background. The details of weathered hands shaping clay are crystal clear.
→ A photorealistic portrait with cinematic lighting and professional composition
2. Stickers & Illustrations
Creating stickers, icons, or digital assets? Be specific about the style and always request a transparent background.
Tips:
- •Clearly state it's a sticker or icon
- •Specify the art style (kawaii, flat design, hand-drawn)
- •Request transparent background
- •Describe colors and outline style
Example Prompt:
A kawaii-style sticker of a smiling avocado wearing tiny sunglasses. Bright cheerful colors with bold black outlines. Cute and bubbly expression. Clean vector-style illustration on transparent background. Perfect for messaging apps.
→ A polished sticker ready to use, with clean edges and transparent background
3. Text Rendering
Gemini excels at putting text in images. Spell out exactly what the text should say, describe the font style, and explain the overall design. Works best with Gemini 3 Pro Image Preview.
Tips:
- •Write the exact text in quotes
- •Describe font characteristics (bold, elegant, playful, modern)
- •Specify text placement and size
- •Mention any decorative elements around the text
Example Prompt:
Design a minimalist coffee shop logo with the text "BREW & CO" in bold, modern sans-serif lettering. The font should feel clean and contemporary. Include a simple steam swirl icon above the text. Monochrome black on white background. Professional and timeless aesthetic.
→ A crisp, professional logo with perfectly rendered text
4. Product Photography
Perfect for e-commerce, advertising, or branding. Create clean, professional product shots that look studio-quality.
Tips:
- •Describe the product and angle clearly
- •Specify background (white, gradient, lifestyle setting)
- •Mention lighting setup (studio lighting, soft shadows)
- •Add styling details (minimalist, luxurious, playful)
Example Prompt:
Professional product photography of a sleek modern water bottle, centered on a pure white background. Soft studio lighting creates a gentle shadow beneath. The bottle's minimalist design and brushed metal cap are clearly visible. Clean, high-end commercial aesthetic perfect for an online store.
→ A magazine-quality product shot ready for your store
5. Backgrounds & Textures
Need backgrounds for websites, presentations, or marketing? Focus on mood, colors, and composition while keeping them suitable for text overlay.
Tips:
- •Describe the overall mood and color palette
- •Specify composition (centered, rule of thirds, minimal)
- •Mention whether it should be busy or simple
- •Consider text readability
Example Prompt:
A minimalist composition featuring a single delicate red maple leaf resting on smooth grey stones. Shallow depth of field with a soft blurred background in warm beige tones. Zen-like and peaceful atmosphere. Plenty of negative space on the right side for text overlay. Clean and sophisticated aesthetic.
→ A beautiful background that won't compete with your text
6. Sequential Art (Comics / Storyboards)
Create panels for visual storytelling with consistent characters and clear scene descriptions. Best used with Gemini 3 Pro Image Preview for text accuracy.
Tips:
- •Describe each panel's composition
- •Maintain character consistency across panels
- •Specify art style (noir, manga, cartoon)
- •Include dialogue or captions if needed
Example Prompt:
Create a three-panel comic in a gritty noir art style with high contrast black and white. Panel 1: A detective's silhouette in a rain-soaked alley, text reads "It was a cold night." Panel 2: Close-up of a mysterious envelope being opened, text reads "Then this arrived." Panel 3: The detective's shocked expression, text reads "Everything changed." Cinematic and dramatic mood throughout.
→ A cohesive comic strip with consistent style and clear storytelling
Image Editing Techniques
Already have an image? Here's how to modify it like a pro
1. Adding & Removing Elements
Provide an image and describe your changes. The model will match the original's style, lighting, and perspective.
- •Describe what to add or remove clearly
- •The model will automatically match the existing style
- •Be specific about placement and size
- •Multiple iterations may refine results
Example:
Add a steaming cup of coffee on the wooden table in the foreground. Make sure it matches the rustic aesthetic of the scene.
2. Masking & Targeted Edits
Use conversational language to define which part of the image to change while keeping everything else intact.
- •Clearly identify the area to modify
- •Describe the desired change
- •Everything else stays the same
- •Works great for fine-tuning specific details
Example:
Change only the sky in this photo to a dramatic sunset with orange and pink clouds. Keep everything else exactly as it is.
3. Style Transfer
Give an image and ask the model to recreate it in a completely different artistic style.
- •Name the target style clearly (watercolor, anime, oil painting)
- •The composition stays similar
- •Artistic interpretation will vary
- •Great for creative exploration
Example:
Recreate this landscape photo as if it were an impressionist oil painting in the style of Monet. Soft brushstrokes and vibrant colors.
4. Multi-Image Composition
Combine elements from multiple images to create something entirely new. Perfect for product mockups or creative collages.
- •Upload multiple reference images
- •Clearly describe how they should combine
- •Specify the final scene or composition
- •Great for visualizing products in new contexts
Example:
Take this product from image 1 and place it in the lifestyle setting from image 2. Make it look natural and professionally integrated.
5. Preserving Key Details
When editing, certain details like faces or logos need to stay perfect. Describe them explicitly in your edit request.
- •Mention what must stay unchanged
- •Be specific about important details
- •Describe those elements in your prompt
- •The model will prioritize preserving them
Example:
Change the background to an urban setting, but keep the person's face, expression, and clothing exactly as they are. No changes to facial features.
6. Sketch to Polished
Upload a rough sketch or doodle and let the model transform it into a finished, professional image.
- •Your sketch doesn't need to be perfect
- •Describe the final style you want
- •Add details about colors, materials, lighting
- •Great for bringing concepts to life
Example:
Transform this rough car sketch into a photorealistic concept car rendering. Sleek modern design with a metallic blue finish. Studio lighting with dramatic shadows. Professional automotive photography quality.
7. Character Consistency
Generate the same character from multiple angles by iterating with different viewpoints. Include previously generated images to maintain consistency.
- •Start with one angle/pose
- •Include that image when requesting new angles
- •Describe the desired new pose or viewpoint
- •Build up a full character sheet gradually
Example:
Show this same character from a 45-degree angle. Keep all the same features - the armor design, colors, and proportions must match exactly.
Advanced Tips from the Pros
Level up your prompts with these expert techniques
→Be Specific
The more detailed you are, the better the AI understands your vision. Instead of "fantasy armor," try: "Ornate elven plate armor etched with silver leaf patterns, featuring a high collar and falcon-wing shoulder guards."
→Provide Context & Intent
Explain what the image is for. The model's understanding of context affects the output. "Design a logo for a high-end minimalist skincare brand" works better than just "design a logo."
→Iterate & Refine
Don't expect perfection on the first try. Use the model's conversational nature to make small changes. Follow up with prompts like: "This is great, but can you make the lighting warmer?" or "Keep everything the same, but make the character's expression more serious."
→Use Step-by-Step Instructions
For complex scenes with many elements, break your prompt into steps: "First, create a background of a serene, misty forest at dawn. Then, add an ancient moss-covered stone altar in the foreground. Finally, place a glowing sword on top of the altar."
→Think Positive (Semantic Negatives)
Instead of saying "no cars," describe what you want: "An empty, desolate street with no signs of traffic." The model understands positive descriptions better than negative ones.
→Control the Camera
Use photography and cinematography language to control composition. Words like "wide-angle shot," "macro shot," or "low-angle perspective" give you precise control over framing.
What Makes Gemini Special
Unique capabilities you won't find elsewhere
High-Resolution Output
Built-in support for 1K, 2K, and 4K images. No need for external upscaling.
Advanced Text Rendering
Creates clear, legible, stylized text perfect for infographics, menus, charts, and marketing materials.
Google Search Grounding
The model can use Google Search to verify facts and generate images based on real-time data like current weather maps, stock charts, or recent events.
Thinking Mode
For complex prompts, the model uses a "thinking" process to reason through your request. It generates temporary "thought images" behind the scenes to optimize composition before delivering the final high-quality output.
Up to 14 Reference Images
Mix and match up to 14 reference images - up to 6 high-fidelity objects to include, and up to 5 portraits for character consistency.
Good to Know: Limitations
Understanding these helps you get better results
**Best languages:** English, Spanish (Mexico), Japanese, Chinese, and Hindi work best. Other languages may have mixed results.
**No audio or video input:** The model works with text and images only.
**Image output count:** The model may not always generate exactly the number of images you request.
**Sweet spot for inputs:** Works best with up to 3 input images. More is possible, but quality may vary.
**Text generation tip:** For images with text, generate the text content first, then ask for an image containing that text. This gives better results.
**SynthID watermark:** All generated images include an invisible SynthID watermark for transparency and attribution.
Ready to Create?
Now that you know the techniques, it's time to put them into practice. Browse our community gallery for inspiration or start creating your own.