How to write AI Prompts for Images: Tips and Examples

You thought creating digital art was hard? Wait till you try writing prompts for AI image generators. You'll feel like you can stare at a blank text box for a week straight trying to figure out the perfect wording, and not just from the "what should I create" paralysis. 

Also, the mind-boggling complexity of these systems, whose ability to interpret your words will only be rivaled by their tendency to completely misunderstand what you're asking for. You'll become the frustrated-keyboard-smashing guy you'd once declared you'd never be.

What is an AI Image Prompt?

It may seem straightforward, but that's hype. It ain't all it's cracked up to be. Fact is: an AI image prompt is the text description you feed to an AI system that tells it what kind of image to generate. Think of it as giving instructions to an artist who's technically brilliant but doesn't quite understand human language the way we do.

These prompts work by connecting with the AI's training data, millions of images paired with descriptions that the system has analyzed. When you type something like "sunset over mountains," the AI searches through its knowledge to find visual patterns associated with those words, then creates a new image based on what it's learned.

The quality of your prompt directly influences what you get back. A weak prompt like "cool picture" will give you something generic and probably disappointing. But a detailed prompt with specific style, composition, and elements will produce something much closer to what you're imagining. The difference can be as dramatic as stick figures versus professional artwork.

For a broader overview of prompt writing fundamentals, check out our Complete Guide to Writing AI Prompts before diving into these specialized tools.

How to Write an AI Prompt for Images

Step 1: Start with a Clear Subject

The foundation of any good prompt is defining exactly what you want to see. Being vague is like telling a taxi driver "take me somewhere nice" instead of giving an actual address.

• Bad example: "A person with a thing"
• Good example: "A weathered old fisherman with a long gray beard holding an antique compass"

Let’s see what this prompt generates with Flixier’s AI image generator:

Source: Flixier

The more specific you are about your main subject, the better the AI can visualize what you're asking for. Think about:

• Who or what is the main focus?
• What are they doing?
• What distinctive features should they have?
• Where are they located?

Step 2: Add Details About Style

After nailing down your subject, you need to tell the AI what artistic style you're going for. This dramatically changes how your image will look.

Some style options include:

• Photorealistic (looks like a photograph)
• Oil painting (textured, traditional art look)
• Watercolor (soft, flowing, transparent)
• 3D render (computer-generated, three-dimensional)
• Cartoon (simplified, exaggerated features)
• Anime (Japanese animation style)
• Pencil sketch (line art, shading)
• Digital art (clean, modern, digital-looking)

For example: "A weathered old fisherman with a long gray beard holding an antique compass, oil painting style with visible brushstrokes"

Source: Flixier

Step 3: Define the Mood, Lighting, and Colors

Not only will your image's mood and atmosphere be affected by these elements, but you might even be OK with spending extra time on this part, as it's what separates amateur-looking results from professional ones.

Specify:

• Lighting conditions (harsh sunlight, soft diffused light, dramatic shadows, backlit)
• Time of day (golden hour, blue hour, midday, night)
• Weather (clear, foggy, rainy, snowy)
• Color palette (vibrant, muted, monochromatic, specific color schemes)
• Emotional tone (peaceful, tense, mysterious, joyful)

Example: "A weathered old fisherman with a long gray beard holding an antique compass, oil painting style with visible brushstrokes, dramatic side lighting at sunset, warm orange and cool blue color palette, melancholic mood"

Source: Flixier

Step 4: Specify Composition and Perspective

How your subject is framed makes a huge difference in the final image:

• Camera angle (eye level, bird's eye view, worm's eye view)
• Shot type (close-up, medium shot, wide shot)
• Composition (rule of thirds, centered, symmetrical)
• Depth (shallow depth of field, everything in focus)

Example: "A weathered old fisherman with a long gray beard holding an antique compass, oil painting style with visible brushstrokes, dramatic side lighting at sunset, warm orange and cool blue color palette, melancholic mood, medium shot from slightly below, rule of thirds composition"

Source: Flixier

Step 5: Add Technical Specifications

For that extra polish, include technical details that elevate the quality:

• Resolution or aspect ratio (16:9, square, portrait)
• Quality terms (highly detailed, 8K, ultra-realistic)
• Rendering specifications (ray tracing, global illumination)
• Camera specifications (35mm, wide-angle lens)

Example: "A weathered old fisherman with a long gray beard holding an antique compass, oil painting style with visible brushstrokes, dramatic side lighting at sunset, warm orange and cool blue color palette, melancholic mood, medium shot from slightly below, rule of thirds composition, highly detailed, 8K resolution"

Source: Flixier

Prompt Writing Tips for Better AI Images

Be Specific, Not Vague

Vague prompts lead to generic, unpredictable results. The more specific you are, the closer the AI will get to your vision.

Vague: "A nice landscape"

Specific: "A misty mountain valley at dawn with pine trees, a winding river reflecting the pink sky, and snow-capped peaks in the distance"

Think about the details that matter to you:

• Time of day (dawn, noon, dusk, night)
• Weather conditions (clear, cloudy, rainy, foggy)
• Specific elements (trees, buildings, people, animals)
• Colors and lighting (warm golden light, cool blue shadows)
• Textures (rough, smooth, metallic, organic)

The difference between "a forest" and "a dense pine forest with morning mist rising between the trees as sunlight filters through the canopy" is enormous in terms of what the AI will generate.

Use Reference Artists or Styles If Needed

Source: Flixier

One powerful technique is referencing known artists or art styles. This gives the AI a clear stylistic direction to follow.

Examples:

• "in the style of Vincent van Gogh" (post-impressionist, swirling brushstrokes)
• "like a Wes Anderson movie scene" (symmetrical, pastel colors)
• "similar to Studio Ghibli animation" (whimsical, detailed backgrounds)
• "inspired by cyberpunk aesthetics" (neon, high-tech, gritty)
• "reminiscent of Art Deco posters" (geometric, bold, elegant)
• "in the manner of Rembrandt" (dramatic lighting, rich shadows)

This works because AI models have been trained on countless artworks with their associated creators and styles. By referencing these, you tap into the AI's understanding of specific visual languages.

For best results, choose artists or styles that:

• Are well-known and widely represented in training data
• Have a distinctive visual style that's easy to recognize
• Match the subject matter you're trying to create

Don't Overload the Prompt with Too Many Elements

Even if you’re the most detail-oriented person in town, and your prompt is the exact opposite of minimalist, studies reveal that AI systems will get confused by too many conflicting elements and can degrade performance. Too many details can cause the AI to prioritize incorrectly or create a messy image.

Source: Flixier

Overloaded: "A futuristic cyberpunk city with flying cars and neon lights and ancient Roman architecture and tropical plants and medieval knights fighting robots and a sunset and rain and snow all at once"

Source: Flixier

Better: "A futuristic cyberpunk city with flying cars and neon lights, with subtle elements of ancient Roman architecture integrated into the skyscrapers"

As a general rule:

• Focus on 3-5 main elements
• Group related concepts together
• Prioritize what's most important
• Consider creating multiple images with different focus areas instead of cramming everything into one

Use Weights to Prioritize Elements

Some AI image generators allow you to assign weights to different parts of your prompt. This tells the AI which elements are most important.

Source: Midjourney
 

For example, in Midjourney you can use colons and numbers:
"A forest landscape::1.3 with small cabin::0.8 and morning mist::1.2"

In Stable Diffusion, you might use parentheses and values:
"A forest landscape (detailed trees:1.3) with (small cabin:0.8) and (morning mist:1.2)"

This would emphasize the detailed trees and morning mist while making the cabin a less dominant element. The exact syntax varies between AI systems, so check the documentation for your specific tool.

Effective weighting strategies:

• Give higher weights (above 1.0) to your main subject
• Give lower weights (below 1.0) to background elements
• Use extreme weights sparingly (below 0.5 or above 1.5)
• Test different weight combinations to find what works best

Consider Negative Prompts

Many AI image generators allow you to specify what you DON'T want in the image. This is called a negative prompt.

For example, if you're creating a portrait and want to avoid common AI issues:
Negative prompt: "deformed hands, extra fingers, bad anatomy, blurry, low quality, text, watermark"

Effective negative prompts can help you:

• Avoid common AI generation problems
• Remove unwanted elements or styles
• Create cleaner, more focused images
• Steer the AI away from inappropriate content

Use Technical Quality Boosters

Certain terms signal to the AI that you want high-quality output:

• "highly detailed"
• "8K resolution"
• "photorealistic"
• "sharp focus"
• "professional photography"
• "masterpiece"
• "award-winning"

These terms act as quality enhancers that tell the AI to put extra effort into rendering details and creating polished images.

Experiment with Aspect Ratios

The shape of your image dramatically affects composition. Most AI generators let you specify aspect ratios:

• 1:1 (square) - good for portraits and symmetrical compositions
• 16:9 (widescreen) - ideal for landscapes and cinematic scenes
• 9:16 (vertical) - perfect for phone wallpapers and character full-body shots
• 3:2 (standard photo) - versatile for most subjects

Choosing the right aspect ratio before generation helps the AI compose the image properly instead of trying to fit your subject into an inappropriate frame.

Learn Model-Specific Terms

Each AI image generator has its own "dialect" of prompt language that works particularly well:

For Midjourney:

• "--stylize" parameter (controls artistic interpretation)
• "--chaos" parameter (adds randomness)
• "--quality" parameter (affects rendering time and detail)

For Stable Diffusion:

• "Unreal Engine" (for video game-like 3D renders)
• "Octane render" (for realistic 3D scenes)
• "trending on ArtStation" (for professional digital art look)

For DALL-E:

• "digital art" (for clean, computer-generated look)
• "photograph" (for realistic images)
• "illustration" (for stylized drawings)

Learning these model-specific terms can significantly improve your results with each particular AI system.

Test and Iterate

A study of novice and expert users of AI image tools yielded quite an interesting find. A startling majority of beginners thought these tools were a piece of cake to use. But they still struggled like hell to get decent images out of them. And strategy for refining prompts? Forget about it. They had none.

The newbies gravitated toward the simplest, most dumbed-down interfaces they could find. They haven't bothered trying different prompt approaches or learning a damn thing from their previous failures. In contrast, the experts were obsessed with image quality above all else. They hunted down tutorials and ran test after test on their prompts to squeeze out better results.

The study showed what you'd expect: beginners were left hanging without the personalized guidance they desperately needed. But the experts leveraged every ounce of their hard-won experience to milk these AI tools for all they were worth. When given a basic prompt to write in a short amount of time, the novice kept trying the same, if it didn't work. The experts wound up testing variations and learning from each attempt.

The most effective prompt writers:

• Start with a simple version of their prompt
• Generate multiple variations
• Note what works and what doesn't
• Refine the prompt based on results
• Keep a prompt journal of successful formulations

Remember that prompt writing is an iterative process. Your first attempt rarely produces exactly what you want, but each generation teaches you something about how the AI interprets your words.

Examples of AI Prompts for Image Generation

Realistic Photography Prompt Example

You thought getting professional-looking photos was hard? Wait till you try this prompt. A well-crafted realistic photography prompt includes technical camera details, lighting specifics, and subject positioning that make the AI generate something that looks like it came from a professional shoot.

Source: Midjourney

"Professional portrait photograph of a 30-year-old Asian woman entrepreneur in a modern office setting, wearing a tailored navy blue suit, confident expression, soft window lighting from the left creating gentle shadows, shallow depth of field with background slightly blurred, shot on Canon 5D Mark IV with 85mm lens at f/2.0, photorealistic, high detail, 4K"

This works because it:

• Clearly defines the subject and their appearance
• Specifies the setting and context
• Includes lighting direction and quality
• Mentions technical photography details that influence the look
• Requests high quality and realism

Fantasy Illustration Prompt Example

Fantasy prompts ain't all they're cracked up to be unless you get specific. Fact is: fantasy prompts need rich descriptive elements that build a cohesive magical world.

Source: Midjourney

"A mystical forest guardian, half-deer half-human druid with antlers and glowing green eyes, wearing robes made of leaves and bark, standing in an ancient enchanted forest with giant mushrooms and floating light orbs, fantasy illustration style, intricate details, magical atmosphere with mist, dappled sunlight filtering through canopy, vibrant colors with emphasis on greens and golds"

This works because it:

• Creates a unique character with specific attributes
• Describes a detailed environment that matches the character
• Specifies the artistic style
• Includes atmospheric elements
• Defines a color palette that enhances the fantasy mood

Abstract Art Prompt Example

Not only will abstract prompts challenge the AI, but you might even be OK with that, as you'll be more eager to see unexpected results. Abstract art prompts should focus on emotions, composition, and artistic technique rather than concrete subjects.

Source: Midjourney

"Abstract representation of human emotions, flowing organic shapes in shades of deep blue and crimson red intertwining on a textured black background, inspired by the style of Wassily Kandinsky, bold brush strokes with visible texture, dynamic composition with movement from lower left to upper right, high contrast, digital painting"

This works because it:

• Provides a conceptual foundation
• Specifies colors and background
• References a known artist for style guidance
• Describes the desired texture and technique
• Indicates composition direction

3D Render Prompt Example

A study of 100 3D artists yielded quite an interesting find. A startling 82 of them reported that specific technical rendering terms dramatically improved their AI-generated 3D images.

Source: Midjourney

"Hyper-realistic 3D render of a futuristic smart home kitchen, minimalist Scandinavian design with white and light wood surfaces, advanced cooking robots, holographic recipe displays, automated food preparation systems, ambient lighting with blue accents, glossy surfaces with accurate reflections, photorealistic textures, 8K, Octane render, global illumination, depth of field"

This works because it:

• Clearly specifies the 3D rendering style
• Describes the setting in detail
• Includes specific design elements
• Mentions lighting and material properties
• References technical rendering terms that improve quality

Tools That Help You Write Better AI Prompts

You thought writing prompts was hard? Wait till you try sorting through all the tools that claim to help. You'll feel like you can browse websites for a week straight trying to figure out which ones actually work, and not just from the "which one is best" paralysis. 

Also, the overwhelming number of options, whose promises to improve your prompts will only be rivaled by their tendency to overcomplicate what should be simple. You'll become the tab-hoarding-browser-crashing person you'd once declared you'd never be.

Prompt Generators

PromptBase

Source: PromptBase

PromptBase offers a marketplace of prompts that have been tested and refined. It ain't all it's cracked up to be, but it's pretty useful. Fact is: it gives you a shortcut. 

Think of it as a recipe book for AI images - someone else did the hard work figuring out what ingredients work together. You can browse by style, model, or subject, see example images, and buy prompts that catch your eye.

Midjourney Discord Community

Source: Discord

Midjourney's Discord Community has channels where users share successful prompts. Not only will you see the exact text they used, but you might even be OK with spending hours scrolling through there, as you'll be more eager to learn from others than to bang your head against the wall trying to figure it out yourself. 

But, there's a slew of new techniques in your life… Some you like, some you don't. And you know what else? Some that work for your style, and some that don't.

DALL-E Discord Community

Source: Discord

DALL-E's Discord Community works similarly. You'll be seeing a lot of shared prompts there, too, because their users are either generous with their discoveries or proud of their results. Either way, you benefit from their experience without having to make all the same mistakes yourself.

Prompt Enhancement Tools

ChatGPT

ChatGPT can help expand basic prompts into more detailed ones. Just tell it your basic idea, and ask it to elaborate with more specific details. It's particularly good at suggesting artistic styles and technical terms that you might not have thought of. 

For example, asking it to "Help me create a detailed prompt for a mountain landscape" will yield suggestions that can significantly improve your results. 

Prompt Craft

Prompt Craft is designed specifically to help refine prompts. It can suggest improvements, add technical details, and help structure your prompts more effectively. 

The tool is particularly useful for beginners who aren't familiar with the specific terminology that tends to work well with AI image generators.

Playground AI

Playground AI offers a visual interface for building prompts. This makes prompt creation more accessible for beginners who might not know all the technical terms. Their interface includes options for adjusting different elements, which is particularly helpful for understanding how to balance your prompt.

Learning Resources

Lexica.art

Source: Lexica

Lexica.art is a search engine for AI-generated images that shows the prompts used to create them. You can search by style, subject, or keyword, making it easy to find examples relevant to your interests. 

With millions of images in their database, it's one of the most comprehensive resources available for studying successful prompts.

PromptHero

Source: PromptHero

PromptHero provides tutorials and guides about prompt engineering. Even if your child's mother is the most tech-savvy person in town, and your learning style the exact opposite of structured, studies reveal that nine out of 10 AI users will see an increase in results quality, which can result in an increase in satisfaction, which can result in setting aside more time for creation, and so on and so on. The result? Some of that precious frustration all to yourself, if not all, dissipates.

CLIP Interrogator

Source: HuggingFace

CLIP Interrogator is a reverse-engineering tool that analyzes existing images and suggests prompts. Know that amazing artwork you saw online? This tool has an uncanny ability to figure out how it might have been created. 

It might not seem as magical once you understand how it works, but the 3 a.m. "aha!" moments when it helps you crack the code on a style you've been trying to replicate can lead to a breakthrough.

Common Mistakes to Avoid When Writing AI Prompts

Being Too Generic

Not only will your generic prompts lead to generic images, but you might even be disappointed with that, as you'll be more eager to create something unique and personal. Generic prompts lead to generic, forgettable images. If you ask for "a beautiful woman," you'll get something that looks like an average of all the "beautiful women" in the AI's training data.

Generic: "A cool car"
Better: "A midnight blue 1967 Ford Mustang Shelby GT500 on a coastal highway at sunset, low angle shot emphasizing power"

The more detailed and distinctive your description, the more unique your image will be. Generic prompts also make it harder for the AI to understand your intent, so it resorts to more random interpretations. 

Common generic terms to avoid:

• "Nice"
• "Beautiful"
• "Cool"
• "Good"
• "Interesting"

Replace these with specific descriptors that actually communicate what you mean by these subjective terms.

Contradicting Descriptors

AI models get confused when you include contradictory elements in your prompt, just like humans would.

Contradictory: "A young elderly woman with long short hair in a bright dark room"
Better: "A woman in her early 70s with a youthful appearance and shoulder-length silver hair in a dimly lit room with one bright spotlight"

Make sure all your descriptors work together logically. If you're not sure, read your prompt out loud and see if it makes sense.

Other common contradictions to avoid:

• Conflicting time periods ("futuristic medieval castle")
• Opposing weather conditions ("sunny rainy day")
• Impossible physical attributes ("tall short person")
• Contradictory emotions ("happy sad expression")

If you want to blend concepts, use clearer language that explains the relationship: "A medieval castle with subtle futuristic elements integrated into its architecture"

Not Matching Style with Subject

Some subjects work better with certain styles. A corporate boardroom scene might look strange in a watercolor style, while a fantasy creature might not work well in a photorealistic style.

Mismatched: "A detailed financial graph in the style of a children's crayon drawing"
Better: "A detailed financial graph in a clean, professional infographic style" or "A playful representation of financial concepts in the style of a children's crayon drawing"

Consider whether your chosen style enhances or detracts from your subject. The style should serve the content, not fight against it.

Effective style-subject pairings:

• Corporate/business subjects + clean, professional styles
• Fantasy creatures + stylized illustration or concept art
• Emotional scenes + expressive painting styles
• Technical objects + 3D rendering or technical illustration
• Natural landscapes + photography or traditional painting styles

When in doubt, think about where you would naturally expect to see this subject in the real world, and what style would typically be used to depict it.

Your Perfect Image Is Just One Prompt Away

Everyone around you will go the distance to convince you that AI will replace human creativity. Not that much anyway. But it won't. AI image generation is just another tool in your creative arsenal, a powerful one that can help you visualize ideas faster than ever before. The human imagination, direction, and taste behind the prompts remain irreplaceable.

Remember that prompt writing is both an art and a science. It requires practice, experimentation, and a willingness to learn from both successes and failures. Start simple, build complexity as you go, and don't be afraid to try new approaches.

The most important thing is to be clear about what you want, specific in how you describe it, and patient as you refine your technique. With time, you'll develop an intuitive sense for how to communicate with AI image generators to create exactly what you envision.

Go ahead and start creating. Your perfect AI image is just the right prompt away.