AI image generation has gone from a niche experiment to a mainstream creative tool in just a few years. Midjourney, DALL-E, Stable Diffusion, and Adobe Firefly are now used by designers, marketers, writers, and hobbyists worldwide. But there's a steep learning curve that most beginners hit immediately: the same tool that produces stunning images for experienced users produces mediocre results for newcomers.
The difference isn't the tool — it's the prompt. This guide teaches you exactly how to write prompts that work, with real examples you can adapt immediately.
What Makes a Good AI Image Prompt?
A good AI image prompt does three things:
- Specifies the subject clearly — What is in the image? Who or what is the main focus?
- Defines the visual style — Is it a photograph, a painting, a digital illustration? What aesthetic?
- Controls the atmosphere — What mood, lighting, and composition should the image have?
Most beginner prompts fail because they only do the first thing. "A cat sitting on a chair" tells the AI what to draw but gives it no guidance on how to draw it. The AI fills in the gaps with its training data defaults — which produces generic, unremarkable results.
Compare these two prompts:
- Weak: "A cat sitting on a chair"
- Strong: "A fluffy orange tabby cat sitting on a vintage wooden chair, warm afternoon sunlight streaming through a window, shallow depth of field, photorealistic, 85mm lens, golden hour lighting"
The second prompt gives the AI specific instructions for subject, environment, lighting, camera settings, and style — resulting in a dramatically better image.
The Anatomy of a Perfect Prompt
Think of a prompt as having six layers. You don't need all six every time, but understanding each one gives you precise control over the output:
1. Subject
The main focus of the image. Be specific. Instead of "a woman", write "a young woman in her 30s with curly red hair". Instead of "a city", write "a futuristic Tokyo skyline at night".
2. Action / Pose
What is the subject doing? "Standing", "running", "looking over her shoulder", "sitting cross-legged". Action adds dynamism and narrative to the image.
3. Environment / Setting
Where is the scene taking place? "In a dense rainforest", "on a rooftop in New York", "inside a cozy library with floor-to-ceiling bookshelves". The environment provides context and depth.
4. Lighting
Lighting is one of the most powerful elements in any image. Specific lighting terms that work well:
- Golden hour — Warm, soft light just after sunrise or before sunset
- Cinematic lighting — Dramatic, high-contrast lighting like a movie scene
- Soft diffused light — Even, shadow-free lighting like an overcast day
- Neon lighting — Colourful artificial light, great for cyberpunk aesthetics
- Rembrandt lighting — Classic portrait lighting with a triangle of light on the cheek
- Backlit / silhouette — Subject lit from behind, creating a dramatic outline
5. Style / Medium
This is where you define the visual language of the image. Options include:
- Photography styles: photorealistic, portrait photography, street photography, macro photography, aerial photography
- Painting styles: oil painting, watercolor, acrylic, gouache, impressionist, expressionist
- Digital art styles: concept art, digital illustration, 3D render, pixel art, vector art
- Cultural styles: anime, manga, Studio Ghibli, Art Nouveau, Art Deco, Bauhaus
- Camera/lens specifications: 35mm film, 85mm lens, wide angle, fisheye, tilt-shift
6. Mood / Atmosphere
The emotional tone of the image. "Melancholic", "joyful", "mysterious", "epic", "serene", "tense". Mood words guide the AI's colour palette, composition choices, and overall feel.
Midjourney vs DALL-E vs Stable Diffusion: Key Differences
Each platform has its own strengths and prompt preferences:
Midjourney
- Best for: Artistic, stylised, and aesthetically polished images
- Prompt style: Responds well to descriptive, evocative language. Loves art style references and mood words.
- Special parameters:
--ar 16:9(aspect ratio),--v 6(version),--style raw(less stylised output) - Weakness: Text in images, precise anatomical accuracy
DALL-E (OpenAI)
- Best for: Photorealistic images, following precise instructions, text in images
- Prompt style: Responds well to clear, literal descriptions. More like giving instructions than painting a picture.
- Special feature: Better at following specific compositional instructions ("place the subject in the left third of the frame")
- Weakness: Can feel less "artistic" than Midjourney for stylised work
Stable Diffusion
- Best for: Maximum customisation, negative prompts, fine-tuned models
- Prompt style: Supports both natural language and comma-separated keyword lists. Negative prompts are a key feature.
- Special feature: Open source — thousands of community-trained models for specific styles
- Weakness: Steeper learning curve, requires more technical knowledge for best results
10 Prompt Templates With Real Examples
Here are 10 ready-to-use prompt templates across different use cases:
1. Portrait Photography
"Portrait of [subject description], [lighting type], [background], shot on [camera/lens], [mood], photorealistic, highly detailed"
Example: "Portrait of an elderly Japanese fisherman with weathered skin and kind eyes, dramatic side lighting, blurred ocean background, shot on 85mm lens, contemplative mood, photorealistic, highly detailed"
2. Fantasy Landscape
"[Fantasy environment], [time of day], [weather/atmosphere], [art style], epic scale, [colour palette]"
Example: "Ancient floating islands covered in lush vegetation, golden sunset, light mist between the islands, digital concept art, epic scale, warm amber and green colour palette"
3. Product Photography
"[Product] on [surface/background], [lighting], studio photography, commercial, clean, [colour scheme]"
Example: "Luxury perfume bottle on a black marble surface, soft studio lighting with subtle reflections, commercial product photography, clean, dark and elegant colour scheme"
4. Character Design
"[Character type], [distinctive features], [clothing/armour], [pose], [art style], full body, white background"
Example: "Female warrior with silver braided hair, ornate golden armour with dragon motifs, battle-ready pose, detailed fantasy illustration, full body, white background"
5. Architecture / Interior
"[Space type], [design style], [materials], [lighting], [atmosphere], architectural photography"
Example: "Minimalist Japanese tea house interior, natural wood and stone materials, soft diffused natural light through shoji screens, serene atmosphere, architectural photography"
6. Food Photography
"[Dish name], [plating style], [surface/background], [lighting], food photography, appetising, [colour palette]"
Example: "Rustic sourdough bread loaf with a cracked crust, on a wooden cutting board with scattered flour, warm natural side lighting, food photography, appetising, warm earthy tones"
7. Abstract Art
"Abstract [concept/emotion], [colour palette], [texture/material], [art movement], [composition style]"
Example: "Abstract representation of solitude, deep blues and purples with flashes of gold, fluid paint textures, expressionist, asymmetric composition"
8. Sci-Fi / Cyberpunk
"[Sci-fi scene], [neon colours], [rain/atmosphere], [time of day], cyberpunk aesthetic, cinematic, [camera angle]"
Example: "Crowded neon-lit street market in a dystopian megacity, rain-slicked pavement reflecting pink and blue neon signs, night time, cyberpunk aesthetic, cinematic, low angle shot"
9. Nature / Wildlife
"[Animal/plant] in [natural habitat], [behaviour/action], [lighting], [season], wildlife photography, [lens type]"
Example: "Snow leopard stalking through deep snow in the Himalayas, intense focused gaze, blue hour twilight, winter, wildlife photography, telephoto lens"
10. Logo / Icon Design
"Minimalist logo for [brand type], [symbol/icon], [colour palette], flat design, vector style, white background"
Example: "Minimalist logo for a coffee brand, stylised coffee bean with steam forming a mountain shape, warm brown and cream colour palette, flat design, vector style, white background"
Common Mistakes That Ruin Your Results
- Being too vague. "A beautiful landscape" gives the AI nothing to work with. Specify the type of landscape, the time of day, the weather, and the style.
- Contradicting yourself. "A dark, bright, colourful, minimalist image" — these instructions conflict. Pick a direction and commit to it.
- Ignoring negative prompts. If you're using Stable Diffusion, negative prompts are essential. Always add: "blurry, low quality, distorted, extra limbs, watermark, text, bad anatomy" to your negative prompt.
- Overloading with adjectives. More adjectives don't always mean better results. "Ultra hyper mega realistic super detailed" is less effective than "photorealistic, 8K, sharp focus".
- Not iterating. Your first prompt rarely produces the perfect image. Treat prompting as a conversation — generate, evaluate, refine, repeat.
- Forgetting composition. Add composition instructions: "rule of thirds", "centered composition", "wide establishing shot", "close-up", "bird's eye view".
Using AI to Enhance Your Prompts
Writing detailed prompts from scratch takes practice. A faster approach is to use an AI prompt enhancer — a tool that takes your basic idea and expands it into a detailed, optimised prompt.
Our AI Prompt Enhancer works like this:
- You enter a basic description: "a cat in a garden"
- The tool expands it into a full prompt: "A fluffy Persian cat sitting among blooming lavender in a sunlit English cottage garden, soft golden afternoon light, shallow depth of field, photorealistic, 50mm lens, peaceful and serene atmosphere"
- You copy the enhanced prompt and paste it into your AI image generator of choice.
If you already have an image you love and want to recreate or build on it, our Image to Prompt tool works in reverse — upload an image and it generates the prompt that would produce a similar result. And for video generation tools like Sora or Runway, our Image to Video Prompt tool creates motion-specific prompts from your images.
The best AI image creators combine their own creative vision with these tools — using AI to handle the technical prompt engineering while they focus on the creative direction.