How to Write AI Image Prompts: 15 Tips with Examples
Start with a Clear Subject, Not a Vague Idea
The single biggest mistake beginners make is writing prompts like "something cool with a dragon". AI image models are literal: they render what you describe, and they fill every gap with a guess. A strong prompt always names one main subject and what it is doing. Compare "a dragon" with "a jade-green dragon coiled around a lighthouse at dusk, sea spray in the air". The second prompt gives the model a subject, a color, an action, a setting and a time of day, which is exactly the information it needs.
A useful habit is to answer four questions before you type anything: who or what is in the image, where are they, what are they doing, and what is the mood? If you can answer those in one sentence, you already have a better prompt than most people ever write. Keep the subject in the first few words of the prompt, because most models weight the beginning of the text more heavily than the end.
Describe the Style Explicitly
If you do not name a style, the model picks one for you, and it will not always pick well. Adding two or three style words transforms a generic output into something intentional. Try suffixes like "digital painting, dramatic lighting", "35mm film photograph, shallow depth of field", or "flat vector illustration, pastel palette". For example: "a cozy mountain cabin in winter, warm light in the windows, watercolor illustration, soft edges".
Styles are also the fastest way to iterate. Keep your subject sentence fixed and swap only the style block: the same cabin as a watercolor, then as an isometric 3D render, then as a moody photograph. Because a generator like PixGenia gives you 20 free images a day with no sign-up, you can afford to burn a few generations just comparing styles before committing to a direction.
Control Lighting and Atmosphere
Lighting is the difference between a flat image and one that feels cinematic. Words like "golden hour", "soft diffused light", "neon glow", "backlit", "volumetric fog" or "harsh midday sun" all push the model toward very different results. A portrait prompted as "soft window light from the left, gentle shadows" will look calm and editorial; the same portrait with "dramatic rim lighting, dark background" will look like a movie poster.
Atmosphere words work the same way: "misty", "rain-soaked", "dusty", "crisp autumn air". Try this full example: "an old bookshop interior, dust motes floating in a beam of morning sunlight, warm amber tones, cinematic photograph". Notice that the lighting description does half of the storytelling. When an image feels boring, changing the light is usually a better fix than changing the subject.
Use Composition and Camera Language
AI models have absorbed millions of photo captions, so photography vocabulary works remarkably well. "Close-up", "wide shot", "aerial view", "low angle", "centered composition", "rule of thirds" and "macro" all reliably shape the framing. For instance: "low angle shot of a skateboarder mid-air against a bright blue sky, wide lens distortion, action photography".
Composition terms are especially valuable when you need images for a specific layout. Need room for text on a banner? Ask for "minimalist composition, subject on the left, large empty sky on the right". Need a phone wallpaper? Ask for "vertical composition, subject in the lower third". Telling the model where things should sit in the frame saves you cropping and editing later.
Add Detail Without Writing a Novel
More words are not automatically better. Past a certain length, models start ignoring or blending instructions, and contradictions creep in. A good target is one to three sentences: subject and action first, then setting, then style and lighting. Every word should earn its place. "Beautiful, amazing, stunning, high quality" adds almost nothing; "weathered leather jacket, silver hair, thoughtful expression" adds a lot.
When you want more detail, be specific rather than superlative. Instead of "a very detailed fantasy castle", write "a fantasy castle with tiered slate roofs, ivy-covered stone walls, and banners snapping in the wind". Concrete nouns beat adjectives every time. If a prompt is getting long, cut the vague words first, and keep the nouns.
Iterate Deliberately: Change One Thing at a Time
Your first generation is a draft, not a verdict. The professionals' trick is to iterate scientifically: change exactly one element per attempt, so you know what caused the difference. If the colors are wrong, adjust only the palette words. If the framing is off, adjust only the composition words. Random rewrites teach you nothing; controlled edits teach you the model's language.
Keep a personal notebook of prompt fragments that worked, like "soft studio lighting, seamless gray background" or "ukiyo-e woodblock print style". After a few weeks you will have a library of reliable building blocks you can recombine. Since PixGenia is completely free, experimentation costs nothing but a minute of your time, so test aggressively and save what works.
Quick Reference: 15 Tips in One List
Here is the condensed checklist. 1) Name one clear subject. 2) Put the subject first. 3) State the action. 4) Set the scene and time of day. 5) Name an explicit style. 6) Describe the lighting. 7) Add atmosphere words. 8) Use camera and composition terms. 9) Prefer concrete nouns over adjectives. 10) Keep prompts to one to three sentences.
And the final five: 11) Specify a color palette when it matters, like "muted earth tones" or "bold complementary colors". 12) Mention the medium — photograph, oil painting, 3D render, pixel art. 13) Leave space for text if the image needs it. 14) Change one variable per iteration. 15) Save every fragment that works. Master these and your hit rate will jump from one good image in ten to one in two or three.