How to Write Image Prompts?

---

Image prompting means describing what you want to see so clearly that the AI can visualize it.
Instead of giving instructions, you describe a scene. What’s there, how it looks and the atmosphere you want.

This guide focuses mainly on Nano Banana and Midjourney, two of the most advanced and popular image models available.

Both can turn your words into stunning visuals, but they work a little differently.

Quick Start Template

Element	Question to Ask Yourself
Subject	Who or what is in the image?
Composition	How is it framed or positioned?
Action	What is happening?
Location	Where does it take place?
Style	What aesthetic or medium fits best?
Lighting	What kind of light or mood?
Color Palette	Warm, cool, muted, bold?
Mood	What feeling should it give?
Aspect Ratio	(Optional) e.g., 16:9, 9:16
Editing Instruction	(Optional, for edits) What should change?

Image generation example

A minimalist living room, wide shot, centered composition, airy loft in Copenhagen, interior design catalog style, soft north light, neutral palette, calm mood.

Image editing example:

Change the sofa color to deep navy blue and add a stack of books on the coffee table.

Core Prompting Principles which can be applied to any model

These principles work across any image generator, including Nano Banana, Midjourne and others.

1. Start with Natural Language

Describe the image as if you were telling someone what to paint. Be clear and specific; avoid vague words like “nice” or “cool.”

2. Build Prompts Step-by-Step

Think in layers: Subject → Environment → Style → Lighting → Mood → Extras

3. Change One Variable at a Time

Adjusting one thing (e.g., lighting or color palette) per run helps you learn how the model responds.

4. Reference Images Are Gold

Whenever possible, pair your prompt with an uploaded image to keep identity, pose, or layout consistent.

5. Use Plain Language for Edits

If the model supports editing (like Nano Banana), be direct:

Replace the background with a cozy living room.
Remove the car in the corner.

Nano Banana Guide

Nano Banana is Google’s next-generation image model. It excels at realism, precision, and consistency, especially across multiple prompts or edits.

Key Capabilities

• Consistent characters: Keep faces, objects, or styles steady across edits.
• Local edits: Modify just one element in plain language.
• Creative composition: Combine different concepts or photos.
• Style transfer: Apply a new look while keeping the same subject.
• Reasoning and sequence logic: Generate multi-step or “what happens next” scenes.

Core Techniques

1. Descriptive Prompting

For realism and detailed control.

A 4K photorealistic portrait of a curly-haired golden retriever mid-jump into a pile of autumn leaves, golden-hour light, shallow depth of field, warm tones.

2. Character Consistency

Preserve identity and key features across images.

1. A whimsical illustration of a glowing mushroom sprite with a bioluminescent cap and vine-like body.
2. Now show the same sprite riding a moss-covered snail through a sunlit meadow.

3. Local, Precise Edits

Conversational editing with clear instructions.

1. Modern living room with gray sofa, wooden coffee table, and potted plant.
2. Change the sofa color to deep navy blue.
3. Add a small vase on the table.

4. Style Transfer

Change aesthetic without altering the subject.

Apply the look of an architectural line drawing to this motorcycle photo.

5. Concept Blending

Combine two or more ideas.

1. Generate a photorealistic astronaut.
2. Generate a rainforest basketball court.
3. Now show the astronaut dunking a basketball in that court.

6. Logical Sequences

Ask for cause-and-effect or storytelling continuity.

1. A chef holding a three-tier cake.
2. Show what happens if they trip.

Nano Banana Tips & Limitations

• Restate defining traits for character consistency
• Use short, clear sentences for edits
• Crop afterward if you need a precise aspect ratio
• Avoid complex text rendering
• Stylization may vary — anchor with adjectives like “cinematic,” “clean lighting,” or “muted colors”

Midjourney Guide

Midjourney specializes in visual style, mood, and composition.
It’s parameter-driven and excels at artistic, cinematic, and conceptual results.

Key Capabilities

• Fine-tuned style control (photography, digital art, film, etc.)
• Parameter prompting for precision
• Fast exploration of aesthetic variations

Core Techniques

1. Style Prompting

Define the artistic look.

Editorial fashion photo on 35 mm film, soft grain, minimal studio set, high-key lighting, neutral palette.

2. Composition and Camera Control

Influence framing and angle.

Wide-angle establishing shot, low-angle perspective, subject centered foreground, city skyline background --ar 16:9

3. Parameter Prompting

Adjust subject emphasis and exclusions.

space station interior::1 astronaut portrait::2 cinematic lighting, volumetric fog --ar 9:16 --no text

4. Weighted Emphasis

Increase focus on one concept by giving it higher weight (::2 or more).
Decrease importance with lower weights.

5. Negative Prompting

Use --no to remove unwanted elements:

--no text, --no people, --no watermark

Midjourney Tips & Limitations

• Use :: weights and --no for balance
• Adjust one parameter per iteration for best control
• Start with a plain descriptive prompt before adding parameters
• No natural-language editing like Nano Banana — you must re-render
• Aspect ratios work reliably only via --ar, not all compositions adapt perfectly

Common Beginner Mistakes to Avoid

Even powerful models struggle when prompts are unclear or overloaded. These are the issues most beginners run into:

• Giving instructions instead of describing a scene (“Make this look cool” → say what you want to see.)
• Using vague adjectives Words like beautiful, nice, cool mean different things to everyone — be specific instead.
• Mixing too many styles at once Stick to one medium or aesthetic per prompt (e.g., film photography or watercolor, not both).
• Writing long paragraphs instead of a concise description Clear, compact prompts outperform rambling ones.
• Changing multiple variables at the same time Adjust one thing per iteration (lighting, style, composition) to understand the model’s behavior.

When should I use which model?

Use Nano Banana when you need precision, editing, and character consistency.
Use Midjourney when you want strong artistic direction, composition control, and fast style exploration.
Both together cover nearly every modern image-generation workflow in 2025.