Step-by-step

🎨 Gemini

Google

📘 Step-by-step 📘 Google · Gemini↗ Gemini · Official sitebeginner 🏠 Everyday life

How to Write a Gemini Image Prompt That Gets Close to What You Pictured

Q: What mistakes should I avoid?

Writing a single vague word. "A beautiful landscape" leaves almost everything open. The fix is to add at least a place, a style and a light source, even in short phrases. Piling on too many ideas at once. Ten objects, three styles and two light sources confuse the result. The fix is to pick one main subject and one clear style, then add details gradually. Expecting a perfect copy on the

Use five simple ingredients (subject, setting, style, lighting and framing) to get images closer to your mental picture on the first try.

6 min read·November 7, 2026

This guide was written by AI. It passed automated fact and quality checks, and our editor spot-checks published guides every day.

See AI World HQ first on GoogleAdd us as a preferred source and our stories show up higher in Google’s Top Stories.Add on Google

You have a picture in your head, but Gemini can only work with the words you type. By the end of this guide, you'll know how to describe any image in five clear parts, so the result looks much more like what you had in mind. This is for anyone who has ever typed "a cat in a hat" and been surprised by the result.

💡 Tip: tap a step’s number when you finish it — a green tick appears and your browser remembers how far you got.

✅ Before you start

A Google account, and access to the Gemini chat in a web browser or the Gemini app on your phone.
Image creation may not be available in every region or on every plan. If you can't find an image option, it may not be switched on for your account yet.
About 15 to 20 minutes. Most of that time is trying, looking and adjusting.

Step 1 of 7

Open Gemini and start a fresh chat

Open Gemini in your browser or app. Start a new chat, usually by tapping a pencil or "new chat" icon, so your image request isn't mixed up with older messages. You'll see an empty text box at the bottom of the screen, often with placeholder words like "Ask Gemini" or "Enter a prompt here." Those words may be different on your device. If you see a button or option for creating images, you can use it, but you can also simply type a request that starts with "Create an image of…" and see what happens. After you send it, Gemini usually replies with one or more pictures, sometimes taking a few seconds.

💬 Try typing thisType your own details into the dashed boxes, then copy."Create an image of a friendly on a ."➤

✅ You'll know it worked when

Gemini replies with a picture, or with a message explaining why it can't make one yet.

Next step ↓

✏️ Sketch — your screen may look a little different

Step 2 of 7

Name the subject

The subject is the main thing in the picture: a who, or a what. Be specific. "A dog" leaves Gemini guessing about breed, size and what it's doing. "A small brown dog with floppy ears, sitting and looking up" gives it real direction. Good subjects usually include one clear action or pose. Keep it to one or two main subjects at first, because crowded scenes are harder to get right. Write your subject as a full phrase near the start of your prompt, so Gemini sees it first. If the subject is a real, named person, skip it: Gemini may not create images of real people, and it's safer to describe a made-up character instead.

💬 Try typing thisType your own details into the dashed boxes, then copy."Create an image of a doing with ."➤

✅ You'll know it worked when

your subject is clearly the first thing you notice in the result.

Next step ↓

✏️ Sketch — your screen may look a little different

Step 3 of 7

Describe the setting

The setting answers where and when. Think of it like a stage: the subject is the actor, and the setting is the backdrop. Name a place and, if it matters, a time of day or season. "In a kitchen" is vague; "in a small sunny kitchen with white tiles" paints a clearer scene. Add one or two background details, such as "with plants on the windowsill," so the scene feels finished. Don't overload it with dozens of objects, since Gemini may add or drop small things anyway.

💬 Try typing thisType your own details into the dashed boxes, then copy."...in a during ."➤

✅ You'll know it worked when

the background matches the place you named, even if the details differ a little.

Next step ↓

Step 4 of 7

Choose a style

The style tells Gemini how the picture should look: a photograph, a watercolor painting, a cartoon, an oil painting or a 3D-style render. This single word can change the whole mood, so it's one of the most useful ingredients. If you want realism, say "photo-like." If you want something playful, say "cartoon style with bold outlines." You can also name a mood, like "calm" or "cheerful." Avoid naming a living artist to copy their exact look; describing the style in your own words is simpler and works just as well.

💬 Try typing thisType your own details into the dashed boxes, then copy."...in a with pale, gentle colors."➤

✅ You'll know it worked when

the result looks like the kind of picture you described, whether that's a painting, a photo or a drawing.

Next step ↓

✏️ Sketch — your screen may look a little different

Step 5 of 7

Add the lighting

Lighting is the difference between a flat picture and one with atmosphere. Name where the light comes from and how it feels: "warm morning sunlight from the left," "soft overcast light," "candlelight," or "a neon glow at night." Lighting also helps set the mood, so it's worth a few words. If you don't mention lighting at all, Gemini will choose something for you, which is fine for quick tests but less predictable when you want a specific look. One clear light source is easier to get right than several.

💬 Try typing thisType your own details into the dashed boxes, then copy."...lit by coming from ."➤

✅ You'll know it worked when

the shadows and brightness in the picture roughly match the light you described.

Next step ↓

Step 6 of 7

Choose the framing

Framing is how the camera sees the scene. Is it a close-up of a face, a wide shot of a landscape, or a view from above? Words like "close-up," "medium shot," "wide shot," "eye level" and "bird's-eye view" are easy to type and make a big difference. If you're not sure, "a medium shot at eye level" is a safe starting point for most subjects. Framing also decides how much of the subject is visible, so say whether you want the whole thing or just part of it.

💬 Try typing thisType your own details into the dashed boxes, then copy."Framed as a from ."➤

✅ You'll know it worked when

the subject's size and position on the screen match the framing you asked for.

Next step ↓

✏️ Sketch — your screen may look a little different

Step 7 of 7

Combine the five parts, send, and refine

Now put everything into one flowing prompt, in this order: subject, setting, style, lighting, framing. Send it, then look at the result carefully. Change one ingredient at a time, such as "make the lighting warmer," so you can see what actually changed. Gemini can usually make a new version when you ask it to adjust, and you can keep going until the picture is close to your idea. It won't read your mind, so small details like exact colors or the number of objects may still differ, and that's normal.

💬 Try typing thisType your own details into the dashed boxes, then copy."Create an image of a with floppy ears, sitting on a porch in a , in a soft watercolor style, lit by warm evening light from the left, framed as a medium shot at eye level."➤

✅ You'll know it worked when

the picture matches most of your five parts, and any remaining differences are things you can fix with one small change.

✏️ Sketch — your screen may look a little different

⚠️ Common mistakes

Writing a single vague word. "A beautiful landscape" leaves almost everything open. The fix is to add at least a place, a style and a light source, even in short phrases.
Piling on too many ideas at once. Ten objects, three styles and two light sources confuse the result. The fix is to pick one main subject and one clear style, then add details gradually.
Expecting a perfect copy on the first try. Gemini works from words, not from your memory of the picture. The fix is to treat the first result as a draft and adjust one part at a time.

🚀 Try it now

Think of one simple picture you've been imagining, such as a bowl of soup by a rainy window. Write one sentence that names the subject, setting, style, lighting and framing, then send it to Gemini. Compare the result with your picture, and change just one word to see how it shifts.

🎯 Next guideHow to Generate Images with Google Gemini: A Beginner's GuideCreate beautiful, custom visuals for work or play in minutes using Google's free conversational AI.

❓ Quick questions

How long does this take?

About 6 minutes — the guide has 7 steps, and you can tick each one off as you go.

Which tool do I need?

This guide uses Google Gemini — but the approach works very similarly in other AI assistants.

Do I need to prepare anything?

A Google account, and access to the Gemini chat in a web browser or the Gemini app on your phone.
Image creation may not be available in every region or on every plan. If you can't find an image option, it may not be switched on for your account yet.
About 15 to 20 minutes. Most of that time is trying, looking and adjusting.

What mistakes should I avoid?

Writing a single vague word. "A beautiful landscape" leaves almost everything open. The fix is to add at least a place, a style and a light source, even in short phrases.
Piling on too many ideas at once. Ten objects, three styles and two light sources confuse the result. The fix is to pick one main subject and one clear style, then add details gradually.
Expecting a perfect copy on the first try. Gemini works from words, not from your memory of the picture. The fix is to treat the first result as a draft and adjust one part at a time.

Keep reading

📘 How to Turn a Selfie Into a Polaroid-Style Portrait With a Gemini Prompt A four-part prompt template (subject, outfit, lighting, frame) so your Gemini-generated image looks like a vintage Polaroid — and still looks like you. 📰 How to Use AI to Turn Your Ideas into Images (Even If You Can't Draw) A simple step-by-step guide to using text-to-image AI tools, plus the prompt-writing trick that gets you the picture you actually wanted. 📘 How to Generate Images with Google Gemini: A Beginner's Guide Create beautiful, custom visuals for work or play in minutes using Google's free conversational AI. 📰 Style References in AI Image Tools: A Plain-English Explainer How showing the AI a picture you like can make a whole set of new images feel like they belong together 📘 Getting started with Google Gemini for everyday life Open Gemini in your browser, sign in with your Google Account, send your first prompt, and learn how to follow up and check the answers. 📘 Turn a casual selfie into a clean LinkedIn headshot with one Gemini prompt A copy-paste prompt template that swaps a phone selfie for a tidy, professional-looking headshot with a simple background.

🔥 See all of this week’s AI trends →

📬 The week’s AI, in your inbox

One friendly email every Sunday — the 5 stories that mattered, in plain English. No spam, unsubscribe anytime.

Was this helpful?

✦ Generated by AI in AI World HQ's automated newsroom, written in plain language. Checked by automated fact and quality gates, and our editor spot-checks published guides every day. Spot a mistake? Use the buttons above.

📘 Prefer it in one file? Photos and Images: The AI Pack (20 guides, free printable PDF)

← Back to all stories