Gemini Prompt Template: Fill-in Formulas for Images and Text

Most bad Gemini results come from prompts that are too short, not from the tool. A template fixes that because it makes you answer the questions Gemini would otherwise guess. Here are the templates I'd keep in a note on your phone, starting with the one Google publishes.

Google's own image formula

Google's Gemini image page gives a starting formula: begin with 'Create' or 'Generate an image of', then the subject, the action and the scene. Its example is a cat napping in a sunbeam on a windowsill. Google then suggests building on it with composition, style, image quality and aspect ratio, and its help centre adds that starting with words like draw, generate or create helps.

Slot	What to write	Example
Starter	Create an image of	Create an image of
Subject	Who or what	a golden retriever puppy
Action	What they're doing	chasing autumn leaves
Scene	Where and when	in a city park in late afternoon
Style	Photo, watercolour, cartoon, charcoal	warm photorealistic style
Composition	Close-up, wide, angle	low angle, puppy in the left third
Aspect ratio	Shape for where it'll go	4:5

Template 1: new image

Copy this prompt

Create an image of {subject} {doing what} {where and when}. Style: {photorealistic / watercolour / cartoon / oil painting}. Composition: {close-up / wide shot / from above}. Lighting: {soft morning / golden hour / studio}. Aspect ratio: {1:1 / 4:5 / 9:16 / 16:9}.

Google's own tip is to be as specific as you can. Its example swaps 'a woman in a red dress' for 'a young woman in a red dress running through a park', and the second one gives Gemini far more to work with.

Template 2: edit your own photo

Google's help centre says editing images isn't available to users under 18, so this one needs an adult account.

Copy this prompt

Use my uploaded photo. Keep {my face, features and skin tone / the product / the layout} exactly the same. Change only {the background / the outfit / the lighting} to {what you want}. Keep the result {photorealistic / natural / in the same style}.

Template 3: combine several photos

Gemini can create a new image based on several uploaded images, so this template names each photo.

Copy this prompt

Use the uploaded photos. Put {the person in the first photo} and {the person or object in the second photo} together in {scene}. Keep every face exactly the same. Match the light on everyone. Style: {photorealistic}.

Template 4: image with text

Copy this prompt

Create a {poster / greeting card / social post} for {occasion}. Main text, spelled exactly: "{your text}". Style: {modern / elegant / playful}, colours: {colours}. Aspect ratio: {4:5}. Leave space at the top for the text.

Always proofread text in an AI image yourself, letter by letter.

Template 5: text tasks

For writing and planning, this four part template does the heavy lifting:

Copy this prompt

I'm {who you are}. I need {the task}. It's for {who will read it}. Please give it to me as {format: a short email / 5 bullet points / a table}, in a {friendly / formal} tone, under {length}.

Copy this prompt

Summarise the uploaded file in {number} bullet points for {who}. Then list anything I need to act on, and anything that's unclear.

Template 6: fix the last answer

Copy this prompt

Keep everything else the same, but change {one thing}.

This tiny template saves the most time. Google's own image page says that if you don't like what you see, you can just ask Gemini to change it, such as the background, an object or an added element.

Template 7: a video or animation idea

If you have a plan that includes video, the same structure works, with movement and camera added. Google's limits page says video uses the most of your allowance, so it pays to get this one complete before you send it.

Copy this prompt

Create a short video of {subject} {doing what} in {scene}. Camera: {slow push in / wide static / follows the subject}. Mood: {calm / joyful / dramatic}. Sound: {ambient sounds / soft music}.

A filled-in example

Here's Template 1 with every slot filled, so you can see how much detail is enough:

Copy this prompt

Create an image of a small bakery owner arranging fresh loaves on a wooden shelf in an early morning shop. Style: warm photorealistic. Composition: medium shot from slightly below. Lighting: soft golden light from the front window. Aspect ratio: 4:5.

Small habits that make templates work

• One topic per chat: Google's limits page says long chats make Gemini re-read the history, which uses more of your allowance. Start a new chat for a new topic.
• Fill every bracket: an empty slot is a guess you've handed to Gemini.
• Save your winners: when a filled-in template works, keep it as your own version.
• Don't paste secrets: keep private details out of any template.

For the wider skill behind these, see the prompt engineering cheat sheet and how to write AI prompts for small business. If you write for a brand, the brand voice prompt template pairs well with Template 5.

My advice

Pick two templates you'll use every week, one image and one text, and fill every slot the first time. You'll notice the jump in quality straight away, and you'll stop burning through your usage on endless 'try again' messages.

Frequently asked questions

What is the best prompt format for Gemini images?

Google suggests starting with Create an image of, then the subject, action and scene, and adding composition, style, quality and aspect ratio.

Can I use templates to edit my own photos in Gemini?

Yes, if your account is 18 or over. Google's help page says editing images isn't available to users under 18.

Why should I start a new chat for each task?

Google says long chats make Gemini re-read the conversation history, which uses more of your usage limit.

Do templates work for text in Gemini too?

Yes. Saying who you are, the task, who it's for and the format you want gives Gemini the context it would otherwise guess.