Free AI image prompts, 24 templates and image to prompt tips (2026)

A good AI image prompt says what appears, where, in what style, framing and light, in what format and what to leave out. This library has 24 ready-to-copy AI image prompt templates for ChatGPT, Gemini and Copilot, from product photos to editing an existing image.

They use the same visual brief we teach in the AI innovation program at Founderz: fewer adjectives, more production decisions. To build a custom prompt in a minute, use the free AI image prompt generator.

Add Founderz as a preferred source

What you’ll take from this

This guide puts the prompts, the method and the differences between tools in one place, so you can go from a loose idea to an AI image you can actually use.

• 24 AI image prompts sorted by tool and by use: product, portrait, ads, illustration, branding, realistic scenes, infographics and editing.
• An image-to-prompt instruction that turns a reference you like into a reusable prompt.
• The eight-field structure behind them, so you can write your own prompt templates without relying on mine.
• How to adapt the same prompt to ChatGPT, Gemini, Copilot, Claude and Midjourney.

ChatGPT prompts for AI image generation

ChatGPT creates and edits images in the same conversation and accepts attached photos, so it covers almost everything on this page. It understands the prompt labels as written. The text in brackets is an example: replace it with your case and delete the brackets. Usage limits depend on your plan, so check them in your account.

Prompts for product and ads

Product photo on a white background for a marketplace

Reference: use the product photo I am attaching. Keep the shape, color, proportions and label exactly the same.
Subject: [a 16.9 fl oz bottle of extra virgin olive oil, dark green glass, cork stopper], front view slightly elevated.
Scene: plain white background for [a marketplace listing].
Style: studio product photography, sharp details and controlled reflections on the glass.
Composition: product centered, filling most of the frame, with a soft contact shadow under the base.
Light: diffused studio light from both sides.
Format: 1:1, high resolution.
Avoid: invented text on the label, logos and extra objects.

Image for a Meta lead generation ad

Subject: [a physical therapist checks their schedule on a phone between two sessions] to advertise [a booking app for physical therapists].
Audience and goal: [independent physical therapists aged 30 to 45]; [lead generation].
Scene: bright, tidy clinic.
Style: natural photography, no stock photo look.
Composition: clean top third for the headline and a free strip at the bottom for the button.
Color: [mint green and white].
Format: vertical 4:5.
Avoid: generated text and logos. I will add the headline later in design.

Prompts for portraits and photo editing

Professional profile photo from your own photo

Reference: use the photo I am attaching. Keep facial features, hairstyle and skin tone without beautifying them.
Subject: professional portrait for [a LinkedIn profile photo], looking at the camera, [warm and confident] expression.
Wardrobe: [a navy blazer over a white t-shirt].
Scene: [neutral gray background with a soft gradient].
Style: studio photography, 85mm lens, natural skin texture.
Composition: tight medium shot with soft background bokeh.
Light: soft from the left.
Format: 1:1.
Avoid: plastic skin, accessories that are not in the photo and any text.

Change the background of a photo

Reference: use the attached photo.
Keep: [the person, their clothing, their posture and the light on their face] and the framing.
Change only: the background to [a bright, out-of-focus office with plants].
Adjust: the light in the background so it matches the direction and temperature of the light on the subject.
Avoid: retouching the face.

Gemini prompts for AI image generation

Gemini accepts images in the conversation and works with your Google documents, so it fits when the image is part of a presentation or a series. These three prompts use that. What your version can generate depends on your account and country.

Lifestyle photo from your product photo

Use the attached product photo as the reference and keep the shape, color and proportions.
Subject: [white running shoes with an orange sole] in use: [a runner ties them on a boardwalk bench].
Scene: sunrise by the sea, blurred surroundings.
Style: natural lifestyle photography, no stock photo look.
Light: golden hour, warm side light.
Color: [orange, light blue and white].
Format: 4:5 for [Instagram].
Avoid: logos from other brands and changes to the shape or color of the product.

Consistent portrait series for a team

I am going to generate portraits of [8] people from [the sales team of a consulting firm]. Keep this system in every image:
Scene: [off-white wall with subtle texture].
Composition: medium shot at eye level, same distance from the camera.
Light: soft from the left with fill on the right.
Style: 85mm lens, natural finish.
Color: [navy and sand].
Format: 4:5.
In each message I will change only the person, based on the reference photo I attach. Do not beautify the features.

Supporting image for a Google Slides slide

Create a supporting image to open a section of a Google Slides presentation about [a logistics company expanding into Portugal].
Subject: [aerial view of a container port at sunset].
Style: sober corporate.
Composition: left half as a clean, dark area for the slide text.
Color: [navy and orange].
Format: 16:9.
Avoid: text, charts and logos.

Copilot prompts for AI image generation

Copilot generates images from the same chat you work in and, with the + button, you can attach an image to modify it. Make the description as specific as you can. These two prompts come from a common office case: images for presentations.

Cover image for a PowerPoint presentation

Create a cover image for a PowerPoint presentation about [office cybersecurity].
Subject: [a tidy desk with a laptop and a brass padlock on the lid].
Style: clean illustration, no stock photo look.
Composition: subject in the right third and two thirds free on the left for the title.
Color: [navy and orange] on a plain background.
Format: 16:9.
Avoid: text, logos and screens with readable content.

Consistent icon set for a presentation

Create [6] icons for a presentation about [office cybersecurity]: password, suspicious email, public wifi, backup, update and padlock.
Style: uniform thin line, same stroke weight, rounded corners.
Composition: 3×2 grid, all the same size.
Color: a single color [navy] on a white background.
Format: 3:2.
Avoid: text.

More AI image prompts: product, ads, illustration, branding and editing

The rest of the library is sorted by use. It works for any AI image generator: paste it as is in ChatGPT, Gemini and Copilot, and the section further down explains the changes for Midjourney. For more photographic direction, see the guide to AI photography prompts.

Product and still life

In e-commerce the image has to match the real product. Attach a photo of the item whenever you can.

Top-down still life for skincare

Subject: [a face cream in an amber glass jar] surrounded by [aloe leaves, cucumber slices and water drops].
Scene: [light travertine stone] surface.
Style: minimalist top-down still life (flat lay).
Composition: symmetrical, product in the center and negative space at the top for the headline.
Light: soft natural light, short shadows.
Color: pastel tones.
Format: 4:5, high resolution.
Avoid: text on the packaging.

Macro of a product detail

Subject: [the dial of an automatic watch with polished steel hands].
Style: macro product photography.
Composition: extreme close-up, focus on [the seconds hand] and progressive blur on the rest.
Light: low-angle studio light from the left to bring out the texture of the metal.
Color: [steel, black and a blue accent].
Format: 16:9.
Avoid: reflections of the photo equipment, dust and invented logos.

Social media and posters

A social image is composed around the text: ask for the free space for the headline and add the text later in design. Always state the format, because each network crops different areas.

YouTube thumbnail

Subject: [a man in his 30s with a surprised expression] looks toward [a cracked piggy bank with coins spilling out], in a video about [how to organize your monthly budget].
Composition: person in the right third; left half free for a three-word headline.
Style: high contrast, plain [bright yellow] background.
Light: frontal studio light.
Format: 16:9.
Avoid: generated text.

Poster with exact text inside the image

Subject: vertical poster for [a jazz concert on a rooftop in Brooklyn on June 14].
Exact headline text: "JAZZ ON THE ROOF". Below, smaller: "June 14, 9:00 PM".
Style: illustration of [a saxophone drawn in ink lines against a sunset sky], large, legible art deco lettering.
Composition: headline at the top, illustration below.
Color: [midnight blue, orange and cream].
Format: 2:3.
Avoid: any text other than the one given.

Illustration, concept art and 3D

In illustration, a clear metaphor works better than a list of styles. In 3D, define material, light and background as in a render.

Editorial illustration for an article

Subject: [a small person sitting on a giant pile of open laptops, with a cloud of notifications above] for an article about [remote work burnout].
Style: [flat digital illustration with grainy paper texture], melancholic mood with a touch of humor.
Composition: free space on the left for the headline.
Color: limited palette [petrol blue, coral and off-white].
Format: horizontal 3:2.
Avoid: text.

3D icon for an app

Subject: [a toy rocket] for [the home screen of a productivity app].
Style: 3D icon in matte plastic with brushed metal details.
Composition: centered object with a contact shadow, plain white background.
Light: main studio light from the top left and a soft glow on the edges.
Format: 1:1.
Avoid: text.

Logos and branding

For visual identity, AI is good for exploring territories before you commit to anything. Redraw the symbol you pick as a vector and review it by hand before registering or printing it.

Visual identity moodboard

Subject: 3×3 grid moodboard for the visual identity of [Brote, a specialty coffee shop in Brooklyn].
Personality: [friendly, sustainable and handcrafted].
Content: textures (kraft paper, ceramic, wood), product photography, interior details and a palette swatch.
Style: natural and bright.
Color: [olive green, terracotta and cream].
Format: 1:1.
Avoid: logos and readable text.

New image in your brand style

Reference: use the attached image as the brand style reference.
Keep fixed: [olive green, terracotta and cream], warm natural light from the left, subtle film grain and negative space on the right.
Subject: [a barista serves a filter coffee at a wooden bar].
Format: 4:5.
Avoid: changing the palette or adding elements from other brands.

Realistic scenes, interiors and infographics

Photorealistic images depend on believable physical details: materials, light direction, lens and straight lines. Words like “4K” or “hyperrealistic” add little; a specific lens and time of day add much more.

Photorealistic interior

Subject: [a 270 sq ft living room in a prewar New York apartment, with high ceilings and original hardwood floors].
Style: [contemporary Mediterranean, light wood, linen and ceramics], interior photography with no render look.
Composition: 24mm wide angle at chest height, straight verticals.
Light: natural morning light through two windows, soft shadows.
Format: 3:2.
Avoid: floating furniture, crooked lines and plastic textures.

Exterior architecture at blue hour

Subject: [a single-family house in concrete and wood on a pine-covered hillside].
Style: photorealistic architecture photography with believable materials: concrete joints, wood grain and real reflections in the glass.
Composition: wide eye-level shot from the garden, 35mm lens, everything sharp.
Light: blue hour, interiors lit with warm light, gradient sky.
Format: 16:9.
Avoid: people and cars.

Step infographic with exact text

Subject: infographic that explains [how to start composting at home] in [4] numbered steps from top to bottom.
Exact text: 1 "Separate food scraps", 2 "Mix with dry material", 3 "Turn it every week", 4 "Use the finished compost".
Style: flat vector, one icon per step and arrows connecting the steps.
Color: [soft greens and browns], cream background.
Format: vertical 2:3.
Avoid: any text other than the one given and spelling mistakes.

Editing images with AI

When editing, separate what is kept from what changes, and use only images you have the right to modify.

Extend the canvas to another format

Reference: use the attached image as the base.
Change: the format from [1:1] to [16:9], extending the scene toward [the left and the right].
Continue: [the sunny beach, the horizon and the sky] with the same light, perspective and grain.
Keep: the original subject without moving or modifying it.
Avoid: new elements that draw attention.

Remove unwanted objects

Reference: work on the attached image.
Remove: [the trash can in the bottom right corner and the cables on the ceiling].
Fill: the area with [the same brick wall and the same tile floor], respecting texture, shadows and perspective.
Keep: everything else, including colors, people and framing.

Change the style of an image

Reference: use the attached photo of [our store seen from the street].
Keep: the composition, the perspective and the position of every element.
Change: the treatment to [a retro 1960s poster illustration, with halftone dots and flat colors].
Color: [tomato red, mustard and cream].
Format: same as the original.
Avoid: text and recognizable faces.

Free image to prompt: turn a reference image into a reusable AI image prompt

Image to prompt means describing an image that already works with the same variables you would use to create it, so you can generate new pieces with the same intent. An image to prompt generator turns any photo into descriptive text, and the best ones let you convert an image into a prompt that works like Midjourney or Flux would expect. It is never a way to copy another brand’s work or identity. First check that you can use the reference: your own image, a licensed one or one approved by your team.

• Ask the model to describe what it sees before it interprets the style.
• Put in brackets what you will change in the next piece and keep the rest fixed.
• Test the prompt without attaching the image. If it keeps the intent, it is reusable.
• Save the prompt, the tool and the format next to the approved image.

Instruction to get the prompt from an image

I am attaching a reference image. Turn it into a reusable prompt for generating AI images.
1. Describe only what is visible, without assuming: subject, action, scene, style, composition, light (direction, hardness and temperature), color, angle, apparent lens, depth of field, texture and aspect ratio.
2. Write the prompt as lines with these labels: Subject, Scene, Style, Composition, Light, Color, Format and Avoid.
3. Put in brackets what would have to change to reuse it, for example [product] or [scene].
4. Add a short one-line version for Midjourney, with the format at the end as an --ar parameter.
5. Intended use: [banner for a skincare store website]. Tool: [Midjourney].
Do not identify real people and do not reproduce brands, logos or signatures that appear in the image.

It works in ChatGPT, Gemini, Copilot or Claude with the image attached. Midjourney also has a Describe function: you upload an image and it returns four text suggestions. Use them as a starting point, because they describe the style but do not know your brief.

How I built these AI image prompts

Every template on this page follows the same structure: subject, scene, style, composition, light, color, format and what to avoid. I use it because each line answers a production decision. If you do not make it, the model makes it for you with the most common choice.

I have been working with prompts since 2020 and 2021, first for ads, web content and team processes, and then with images. With images I learned the same thing as with text: if you talk to the AI as in a chat, it gives you the most generic result. Adjectives are rarely missing. The brief is: which piece it is, for which channel and what it has to leave free.

An old-school lesson: in 1990 I coded teletext pages on BBC computers, and the rule was WYSIWYG, what you see is what you get. An image generation model works the same way, which is why this kind of prompt engineering has more to do with briefing than with vocabulary.

The generator turns your answers into a prompt, so you can use the AI image generator in a few seconds and paste the result into any tool. It is completely free to use, and the prompt in seconds is meant to be edited, not accepted blindly. Use our AI image prompt generator when no template fits, and use the templates when you already know the use case.

Order matters too: I put the subject first, because it is the part that is not negotiable, and the constraints last, separated, so you can review them at a glance.

How the prompt changes in ChatGPT, Gemini, Copilot, Claude and Midjourney

• ChatGPT: labels as written. Choose the format in the aspect ratio selector or write it in the prompt. To edit, select the area and describe the change, and ask for one adjustment per message.
• Gemini: accepts photos and documents. Put what must stay the same first and repeat that instruction if you chain several edits.
• Copilot: be as specific as you can and attach the image with the + button if you want to modify it. Ask for the format in words.
• Claude: does not generate images. Use it to write or improve the prompt, or to turn a reference into a prompt, and then paste it into the image tool.
• Midjourney: prefers a shorter version and the parameters at the end: –ar 4:5 for the format, –no for exclusions and –sref to lock a style.

This is the product-on-white prompt adapted to Midjourney:

product photo of [a dark green glass bottle of extra virgin olive oil with a cork stopper], front view slightly elevated, plain white background, soft diffused studio light, sharp details, soft contact shadow --ar 1:1 --no text, logos, extra objects

In Midjourney and Stable Diffusion, if the result ignores details, try a shorter version like this one. In Stable Diffusion, exclusions go in the negative prompt field, separate from the description.

Which AI image generator and image model should you choose for free AI image creation?

AI image generation no longer depends on a single tool, and the best AI image generator is the one that fits your use case. AI art and visuals with AI now come from many tools, so choose by the result you need. In ChatGPT, images come from OpenAI’s GPT Image models (DALL·E before that). Whatever you pick, the result depends more on the text prompt than on the tool.

Tools like Flux, Stable Diffusion and Midjourney are different AI models, and each reads the same text in its own way. Major AI models differ in typography, photorealism and 3D renders, so test the same prompt in two of them before you commit. For photorealism, ask for a specific lens and light. For text effects and typography, write the exact words in quotes. If you are new to this, start with ChatGPT or Copilot, which let you iterate in the same conversation, and move to Midjourney or Firefly when you need one style across many pieces. Use these ready-to-use templates as the base of your creative process, and optimize one variable at a time for creative efficiency and highly accurate results. Many AI tools can export a PNG, so state the format first.

Any major AI chat also works as an image to prompt tool. To turn any image into a detailed AI prompt, convert the image into labeled lines, as in the instruction above, then reuse that detailed prompt to generate prompts for new pieces. The same method works as a free image to prompt workflow with the LLMs you already use: the text prompt describes the image, and the image model generates the new one. Copy the prompt, change the brackets and use it to generate images with any of the AI image models above.

How does an AI image prompt generator work with generative models?

A prompt generator takes a short idea or a reference picture and expands it into detailed text prompts with subject, style, light, framing and format. You can use it to generate prompts for any image model, then paste the result into GPT Image, Midjourney, Flux or Stable Diffusion. You can also write the same structure by hand with the templates above, which is free and gives you more control over the image generation.

Every free AI image prompt generator does the same thing: image-to-text analysis, then a detailed prompt you can edit. A good image prompt generator also lets you ask for image variations, such as a new angle, palette or aspect ratio, without rewriting the brief. Keep a prompt library of the images that worked, so each detailed prompt becomes a template for e-commerce, ads or social posts. Whether you use an AI prompt generator or write by hand, the generator is only as good as your brief.

Which AI models work with these free AI image prompts?

These templates work in ChatGPT, Gemini, Copilot, Midjourney, Flux and Stable Diffusion, because each AI model reads a text prompt with subject, style, light and format. Change only the model syntax, such as the aspect ratio flag in Midjourney, and keep the rest.

To turn any photo into a prompt, use the image to prompt method above. To generate prompts from scratch, fill the brackets in a template and run it in your image generation tool of choice.

What mistakes ruin an AI image prompt?

Four failures show up most: asking for a style without a brief, piling up empty adjectives, mixing several changes in one message and not reserving room for text. All four have a simple fix.

• Asking for a style without a brief: describe the piece before the aesthetic. “4:5 Instagram ad with room for the headline” guides more than “modern style”.
• Piling up empty adjectives like “8K” or “epic”: swap adjectives for decisions, such as an 85mm lens or soft side window light.
• Putting several changes in one message: ask for one change per message when editing and repeat what stays.
• Not reserving room for text or reviewing the result: check hands, text, reflections and product details before you publish, because that is where any AI fails most.

A five-step workflow to generate prompts and AI images

A good creative workflow treats each generated image as a draft of the brief, not as the final image. Generating a prompt is only the first step, and perfect prompts come from variations. These are the five steps I follow with any AI image generator, whether it is ChatGPT, Copilot, Midjourney or Stable Diffusion.

1. Define what the image is for and where it will be seen, because that sets the format and aspect ratio (1:1, 4:5, 9:16 or 16:9).
2. Write the prompt with one line per decision, or build it with the generator, and attach the reference image if you have one.
3. Generate a few versions and keep the one that best solves the brief, not the flashiest.
4. Fix it in parts: one change per message (the background, the light, an object) instead of rewriting the whole prompt.
5. Before you publish, check the commercial use terms of the tool and plan you use, and save the prompt that worked.

The same workflow works for AI portraits and profile photos, product images, thumbnails, social media posts and illustrations. For a series, lock the style, natural light, color palette and framing, and change only the scene or the person. With that system, the prompt library becomes your own template for any AI image generation task.

Can you use images made with AI commercially?

It depends on each tool’s terms and on how you use it. Before you publish, read the service’s terms of use, do not reproduce third-party brands, characters or protected works, and do not generate images of real people without their permission.

Copyright rules differ by country, and an image generated without enough human creative input may not be protected as your own work. If the image will be a central piece of your brand, talk to an intellectual property specialist.

Frequently asked questions about AI image prompts

These are the most common questions when you start writing prompts to create images, with the answer first and the library above to copy the examples from.

How do I write a prompt for an image?

Write one line per decision: what appears, where it is, in what style, how it is framed, with what light and colors, in what format and what must stay out. Add what the image is for. If the result does not fit, change a single one of those lines and generate again.

What are the best ChatGPT prompts for creating images?

The ones that give the full brief in natural language. ChatGPT understands the labels in this library as written and lets you iterate in the same conversation, so ask for one adjustment per message (“now change only the background”) instead of rewriting the whole prompt each time.

How do I get a prompt from an image (image to prompt)?

Attach the image to ChatGPT, Gemini, Copilot or Claude and ask it to describe only what it sees and organize it into subject, scene, style, composition, light, color, format and avoid. You have the full instruction in the image-to-prompt section. Use it only with your own images, licensed ones or ones approved by your team.

Is there a free tool that writes AI image prompts for you?

Yes, the Founderz generator is free. It asks what you want to create, the channel, the style and the constraints, and arranges your answers into a structured prompt you can paste into your image tool.

What prompt should I use for realistic photos?

Add camera decisions to the base structure: lens (35, 50 or 85mm), angle, depth of field and type of light, for example “medium shot, 85mm lens, soft natural side window light, blurred background, 4:5 format”. The portrait and realistic scene prompts above follow that pattern.

Should I write the prompt in English or in another language?

English is the safe default, and ChatGPT, Copilot and Adobe Firefly also handle other languages well. In Midjourney and Stable Diffusion, if the result ignores details, try a shorter English version. The image-to-prompt instruction helps you get that short version.

Can I use Claude to generate images?

Claude does not generate images. It does help you write the prompt, improve it and turn a reference image into a reusable prompt. Then you paste it into ChatGPT, Gemini, Copilot or Midjourney.

What is the best free AI image creator?

To start for free, ChatGPT and Microsoft Copilot create images from the chat. If you want logos or icons, Recraft lets you generate and export SVG with its free account. None is the best for everything: choose by use and check the limits and commercial terms of each plan.

Your next step with AI image prompts

Pick a category, copy a template and change only what is in brackets. When an image works, save the prompt and note which variable made the difference. You will find prompts for other areas in the Founderz AI prompt library.

If you want to apply this method to design, advertising or content projects with Midjourney, Adobe Firefly or Recraft, the AI innovation program at Founderz teaches it with real workflows.