How to Write AI Image Prompts: A Practical Guide with Examples
By the Shotari Team ยท
The difference between a frustrating AI image session and getting exactly what you pictured comes down to one skill: writing a good prompt. Here's a simple, repeatable formula โ plus copy-paste templates you can use right now.
May 20, 2026 ยท 8 min read
On this page
The 5-Part Prompt Formula
Every strong image prompt answers five questions, in roughly this order: Subject โ Style โ Composition โ Lighting โ Details. You don't need all five every time, but naming the subject and style alone will dramatically improve your results. Think of it as describing the photo to someone who will shoot it for you.
Full example
"Photorealistic product photo of a glass perfume bottle [subject], minimalist studio style [style], centered with negative space [composition], soft diffused lighting from the left [lighting], pale pink background, sharp focus, high detail [details]."
1. Subject
The most important part. State exactly what should be in the frame โ the object, person, or scene. Be concrete: not "a drink" but "a glass of iced coffee with condensation on a marble counter."
โ Good
a ceramic coffee mug on a wooden desk
โ Weak
something cozy
2. Style
Define the visual language: photorealistic, watercolor, 3D render, flat illustration, anime, film photography. This single word changes everything about the output.
โ Good
photorealistic product photography
โ Weak
(no style stated)
3. Composition
Tell the AI how to frame the shot: close-up, wide shot, top-down/flat-lay, eye-level, rule of thirds, centered. Add aspect ratio context if it matters.
โ Good
top-down flat-lay, centered, generous negative space
โ Weak
nice angle
4. Lighting & Mood
Lighting defines the emotional tone. Soft natural light, golden hour, studio softbox, dramatic side light, neon glow. Pair with a mood word: warm, moody, clean, energetic.
โ Good
soft morning window light, warm and calm
โ Weak
good lighting
5. Details & Quality
Finish with the small specifics and quality cues: colors, materials, textures, and terms like "sharp focus," "high detail," "4K." These refine the last 10% of the result.
โ Good
muted earth tones, visible ceramic texture, sharp focus
โ Weak
make it pretty
Copy-Paste Templates
Swap the [BRACKETS]for your subject and paste straight into Shotari's prompt box.
Ecommerce product
Photorealistic product photo of [PRODUCT] on a [SURFACE], studio softbox lighting, clean white background, centered composition, sharp focus, high detail, commercial quality.
Social media graphic
[SUBJECT] in a flat illustration style, bold flat colors, centered with negative space for text, energetic and modern mood, clean vector look.
Lifestyle / brand
[SUBJECT] in a natural lifestyle setting, photorealistic, soft golden-hour window light, shallow depth of field, warm and inviting mood, authentic feel.
Portrait / headshot
Professional headshot of [PERSON], photorealistic, soft studio lighting, neutral gray background, eye-level, sharp focus on the eyes, natural skin texture.
Concept / hero image
[SCENE], cinematic photography, dramatic side lighting, wide shot with rule-of-thirds composition, moody atmosphere, rich color grading, 4K detail.
Prompts for Editing (Not Generating)
When you upload a photo to edit, the rules flip: the AI already sees the image, so only describe the change. Don't re-describe the whole scene โ that confuses the model and risks regenerating everything.
Remove an object
Remove the person on the left side of the photo and fill the area naturally to match the background.
Change background
Replace the background with a clean light-gray studio backdrop, keeping the subject and its edges natural.
Adjust lighting
Brighten the underexposed face and add soft, even lighting while keeping the rest of the photo unchanged.
Restyle
Convert this photo into a Ghibli-inspired illustration style while keeping the composition and subject the same.
Background swaps are the most common edit of all โ we covered them in depth, with a real before/after test, in how to remove the background from a photo.
Common Mistakes
- โ
Vague subjects
"Make a nice image" gives the AI nothing to work with. Always name the concrete subject first.
- โ
Stacking 20 adjectives
Too many competing descriptors confuse the model. 5โ8 well-chosen words beat 20 random ones.
- โ
No style direction
Without a style word, you get the model's default look โ usually not what you wanted. Always state photorealistic vs illustration vs 3D.
- โ
Negatives the model ignores
"No text, no people" works inconsistently. Better to describe what you DO want filling that space.
- โ
Editing prompts that re-describe everything
When editing an uploaded photo, only describe the change โ not the whole scene. The AI already sees the image.
Conclusion
Good prompting is a skill you build in an afternoon. Start with the 5-part formula, lean on our ready-made prompt templates, and pay attention to which words actually changed your result. Within a dozen generations you'll develop an instinct for what works.
The best way to learn is to iterate. With 10 free daily credits on Shotari, you can test the same idea with five different prompt structures and see exactly what each word does.
Try these prompts now โ
Open the AI Image Generator, paste a template, swap the brackets, and generate. Free accounts get 10 credits daily.
Open AI Image Generator