> ## Content Index > Fetch the complete content index at: https://www.hyperank.ai/blog/llms.txt > Use this file to discover other available public pages before exploring further. # How to Write AI Image Prompts (Free Templates and Examples) - URL: https://www.hyperank.ai/blog/ai-image-prompts/ - Published: 2026-10-24T19:30:05.000Z - Updated: 2026-10-24T19:30:05.000Z - Description: Discover how to effectively use an ai image prompt generator with our step-by-step guide featuring free templates and examples for your creative projects. - Author: Hyperank AI *ai image prompt generator | Updated October 2026 | hyperank Editorial Team | 30 minutes to learn, 1-2 weeks of practice | Beginner* ## What You'll Learn **To write a good AI image prompt, describe five things in plain language: the subject and what it's doing, the style or medium, the mood, the lighting and camera view, and the composition or format. Then add what to leave out as a negative prompt, generate a few versions, and change one element at a time.** The formula works in ChatGPT, Midjourney, Adobe Firefly, Google's Imagen and Stable Diffusion, and does what people expect from an ai image prompt generator: turns a vague idea into a specific, repeatable instruction. This guide, How to Write AI Image Prompts (Free Templates and Examples), walks through each part with copy-ready templates. - **A reusable formula:** One structure you can apply to any image, from product shots to blog headers. - **Copy-ready templates:** Fill-in-the-blank prompts for photos, illustrations, product images and social posts. - **Better control:** How lighting, camera terms and composition words change the result. - **Fewer bad generations:** How to use negative prompts and fix the most common problems. **Prerequisites:** Access to any AI image generator (most have a free tier) and a clear idea of where the image will be used. --- ## Why Writing Good AI Image Prompts Matters in 2026 AI image generation went mainstream fast. When OpenAI upgraded image generation inside ChatGPT in March 2025, [more than 130 million users generated over 700 million images in the first week](https://techcrunch.com/2025/04/03/chatgpt-users-have-generated-over-700m-images-since-last-week-openai-says/?ref=hyperank.ai), according to OpenAI's Brad Lightcap, as reported by TechCrunch. Marketing teams now use them daily for headers, ad concepts, mockups and social posts. The tools are capable, but the output still depends on the instruction. A prompt like "a coffee shop" gives you a generic stock-photo look. A prompt that names the subject, setting, light, lens and format gives you something you can actually publish. The difference is rarely the model; it's the prompt. For brand teams, a written formula also keeps images consistent when several people are generating them. --- ## The Process at a Glance | Step | Action | Time | Outcome | | ---- | ---------------------------------------- | ----- | ------------------------------------ | | 1 | Learn the five-part prompt formula | 5 min | A structure for every prompt | | 2 | Understand how image models read prompts | 5 min | Know what words actually change | | 3 | Define the subject and action | 3 min | A clear focal point | | 4 | Choose style, medium and mood | 3 min | A consistent visual look | | 5 | Set lighting, camera and composition | 5 min | Images that look intentional | | 6 | Copy and adapt a template | 2 min | A fast, repeatable starting point | | 7 | Write prompts for social media visuals | 5 min | Platform-ready images in right sizes | | 8 | Add negative prompts to remove problems | 2 min | Cleaner, more usable results | **Total time:** About 30 minutes to learn the method, then one to two weeks of regular practice to write strong prompts quickly. --- ## Step 1: Use the Image Prompt Formula ### What You're Doing Most bad AI images come from prompts that skip half the picture. You're adopting one fixed structure so every prompt covers the details an image model needs, in the same order, every time. ### How to Do It 1. Write the **subject and action** first: who or what is in the image, and what is happening. 2. Add the **setting**: where it takes place and what's in the background. 3. Name the **style or medium**: photograph, watercolor, 3D render, flat vector illustration. 4. Describe the **lighting and mood**: soft morning light, moody and cinematic, bright and cheerful. 5. Specify the **camera and composition**: close-up, overhead shot, wide angle, subject on the left third, and the aspect ratio. Google's [Imagen prompt guide](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/image/img-gen-prompt-guide?ref=hyperank.ai) describes the same core idea: start with subject, context and style, then refine with photography modifiers such as camera proximity, lens type and lighting. ### Example | Part | Weak | Strong | | ---------------------- | ------------- | ----------------------------------------------------- | | Subject and action | A woman | A woman in her 30s reviewing a laptop dashboard | | Setting | Office | Bright co-working space with plants in the background | | Style | Nice | Editorial photograph | | Lighting and mood | Good lighting | Soft natural window light, calm and focused | | Camera and composition | None | Medium shot, 35mm lens, subject on left third, 16:9 | ### What Done Looks Like You can write a one- to three-sentence prompt that covers all five parts without having to think about the order. --- ## Step 2: Understand How AI Image Prompts Work ### What You're Doing You're learning how image models turn text into pictures, so you know which words carry weight and which are ignored. ![Business details panel with a Complete with AI button beside the website field](https://amhvrlzbdnkjotzniuhu.supabase.co/storage/v1/object/public/hyperank/screenshots/guide-4417e527-bc61-4fa0-9727-76587941dab9.webp) ### How to Do It 1. **Know the basic mechanism:** Most current generators are trained on huge sets of images paired with captions. They learn which visual patterns go with which words, then build a new image that matches your description. 2. **Write concrete nouns and adjectives:** "Golden retriever puppy on a red wool blanket" gives the model clear visual targets. "A happy, cozy feeling" gives it very little. 3. **Put the important details first:** Many tools give earlier words more influence, so lead with the subject. 4. **Match the tool's style of input:** Conversational tools like ChatGPT handle full sentences and follow-up edits ("make the background darker"). Midjourney and Stable Diffusion respond well to descriptive phrases and support parameters such as aspect ratio. ### What Done Looks Like You stop writing abstract goals and start writing visual descriptions a camera or illustrator could follow. --- ## Step 3: Define the Subject and Action ### What You're Doing With the formula and the mechanics in hand, start where the model starts: the subject. You're making it unmistakable, because it's the part the model gets right or wrong first. ### How to Do It 1. Name the subject specifically: not "a car" but "a matte black electric SUV". 2. Add two or three defining details: age, clothing, material, color, size. 3. Give it an action or state: "pouring coffee", "mid-jump", "resting on a marble counter". 4. Limit the main subjects to one or two. More than that and models often blend or drop details. ### Common Mistakes - **Too many subjects:** "A chef, a waiter, three customers and a dog" usually produces a confused scene. Pick the hero and keep others in the background. - **Relying on text in the image:** Newer models handle short words better than older ones, but long text and logos are still unreliable. Add text and logos afterwards in a design tool. (Anyone who has asked for a simple "SALE" banner and received "SAEL" knows the feeling.) ### What Done Looks Like Someone reading only your first sentence could sketch the main subject and what it's doing. --- ## Step 4: Choose Style, Medium and Mood ### What You're Doing Once the subject is clear, the next question is what kind of picture it should be. You're choosing that deliberately, so the model doesn't default to a generic look. ### How to Do It 1. Pick the medium: photograph, oil painting, watercolor, pencil sketch, 3D render, flat vector, isometric illustration, claymation. 2. Add a style reference in general terms: "mid-century modern poster", "Scandinavian minimalist", "1990s film photo". Avoid naming living artists, which raises copyright and ethics concerns and is blocked by some tools. 3. Name the mood with one or two words: calm, energetic, nostalgic, premium, playful. 4. Add a color palette if brand matters: "muted navy and sand tones", "high-contrast black and yellow". ### Example | Use case | Medium and style | Mood and palette | | ------------------- | ----------------------------------- | ----------------------------------------- | | SaaS blog header | Flat vector illustration, isometric | Clean, optimistic; blue and white | | Premium product ad | Studio product photograph | Luxurious; black background, gold accents | | Wellness brand post | Soft film photograph | Calm; warm beige and sage green | | Event announcement | Bold graphic poster style | Energetic; high-contrast brand colors | ### What Done Looks Like Your prompt clearly states one medium, one mood and, where needed, a palette, and repeated generations share the same look. --- ## Step 5: Set Lighting, Camera and Composition ### What You're Doing You're borrowing a photographer's vocabulary to control how the image looks, not just what's in it. ### How to Do It 1. **Lighting:** Choose one. "Soft natural window light", "golden hour sunlight", "dramatic side lighting", "bright even studio lighting", "neon glow". 2. **Camera distance:** Extreme close-up, close-up, medium shot, full-body, wide shot. 3. **Angle:** Eye level, low angle (makes the subject look powerful), overhead flat lay, bird's-eye view. 4. **Lens and focus:** "35mm lens" for natural scenes, "85mm portrait lens" for flattering faces, "macro lens" for details, "shallow depth of field" for a blurred background. 5. **Composition:** "Subject on the left third", "centered and symmetrical", "plenty of empty space on the right for text". ### What Done Looks Like Your images look composed and intentional, and you can predict how a change in lens or angle will affect the result. --- ## Step 6: Copy and Adapt These Templates ### What You're Doing You're using proven fill-in-the-blank prompts so you don't start from zero, and so your team produces consistent images. ![Brand Hub dashboard showing setup cards and Save everything button](https://amhvrlzbdnkjotzniuhu.supabase.co/storage/v1/object/public/hyperank/screenshots/guide-e57c8001-1885-441e-b57c-e74ff342a8bb.webp) ### How to Do It 1. Pick the template closest to your use case. 2. Replace every bracket with a specific detail. 3. Generate three to four versions and keep the best. 4. Save the final prompt in a shared document so others can reuse it. ### Example | Template | Prompt | | ----------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Lifestyle photo | A \[person\] \[action\] in \[setting\], editorial photograph, \[lighting\], \[mood\], medium shot, 35mm lens, shallow depth of field, \[aspect ratio\] | | Product shot | A \[product, material, color\] on \[surface\], studio product photography, \[background color\] background, soft diffused lighting, centered, sharp focus, \[aspect ratio\] | | Blog illustration | Flat vector illustration of \[concept as a scene\], \[palette\], clean lines, minimal detail, empty space on the \[side\] for a headline, 16:9 | | Flat lay | Overhead flat lay of \[items\] on \[surface\], \[palette\], soft natural light, neatly arranged, 1:1 | Filled in, the product template might read: "A matte white ceramic coffee mug on a light oak table, studio product photography, sage green background, soft diffused lighting, centered, sharp focus, 4:5." ### What Done Looks Like You have a small, saved library of templates and can produce a usable first draft image in under two minutes. --- ## Step 7: Write Prompts for Social Media Visuals ### What You're Doing Templates get you a good image; social platforms need the right shape too. You're tailoring prompts to each format, so images don't get cropped badly and feel native to the feed. ![Media preferences panel showing internal links and an Add CTA toggle](https://amhvrlzbdnkjotzniuhu.supabase.co/storage/v1/object/public/hyperank/screenshots/guide-48d0a451-f28e-4550-a21e-2f5f8b4d01e4.webp) ### How to Do It 1. Start with the aspect ratio for the placement (see the table below). In Midjourney, add it as a parameter such as `--ar 4:5`; in ChatGPT or Firefly, state it in the prompt or choose it in the settings. 2. Keep the subject large and simple. Social images are viewed small on phones. 3. Ask for empty space where captions, logos or text overlays will go. 4. Use your brand palette consistently across posts. 5. Write the caption separately. Our free [Instagram caption generator](https://www.hyperank.ai/tools/instagram/instagram-caption-generator?ref=hyperank.ai) and other [Instagram tools](https://www.hyperank.ai/tools/instagram?ref=hyperank.ai) help with the words that go alongside the image. ### Example | Placement | Aspect ratio | Prompt tip | | -------------------------------------------- | ------------ | ------------------------------------------------------ | | Instagram feed (portrait) | 4:5 | Subject centered, bold colors, little fine detail | | Instagram and Facebook Stories, Reels covers | 9:16 | Keep the top and bottom clear for interface elements | | LinkedIn shared image | 1.91:1 | Professional setting, space for a headline | | X (Twitter) post image | 16:9 | Wide scene, subject clear when cropped in the timeline | ### What Done Looks Like Each image is generated at the right shape for its placement and leaves room for your text and logo. --- ## Step 8: Decide What to Leave Out With Negative Prompts ### What You're Doing Even a strong prompt sometimes produces the same annoying flaw again and again. Here you tell the model which unwanted elements to avoid, such as extra fingers, blur, watermarks or clutter. ![Competitors panel listing an added competitor with a Continue button](https://amhvrlzbdnkjotzniuhu.supabase.co/storage/v1/object/public/hyperank/screenshots/guide-3c3dbc0c-cc05-4095-a547-4ebf6bad009d.webp) ### How to Do It 1. **Midjourney:** Add the `--no` parameter, for example `--no text, watermark, people`. 2. **Stable Diffusion and many open tools:** Use the separate negative prompt field, for example "blurry, low resolution, extra fingers, distorted hands, watermark". 3. **Google Imagen:** List the unwanted items plainly in the negative prompt. Google's guide advises against instructive words like "no" or "don't": write "wall, frame", not "don't show walls". 4. **ChatGPT and other conversational tools:** There's usually no separate field, so describe what you want instead: "plain white background" works better than "no busy background". Ask for edits in a follow-up message. ### Common Mistakes - **Mentioning the unwanted thing in the main prompt:** "A beach without people" can add people, because "people" is now in the prompt. Describe "an empty beach" instead. ### What Done Looks Like Repeated problems such as clutter, blur or stray text disappear from your generations without changing the main subject. --- ## What to Do After You've Learned the Formula - **Phase 1, week one, practice daily:** Write five prompts a day for real projects. Change one element at a time (lighting, lens, style) and note what changed. - **Phase 2, weeks two to four, build a brand prompt library:** Save your best prompts by use case, with the tool, settings and final image. Agree on a standard style line your whole team adds, such as "soft natural light, muted navy and sand palette". - **Phase 3, ongoing, refine and review:** Check that images are accurate (no misleading product details), follow each tool's usage terms, and that AI-generated visuals are labeled where your platform or policy requires it. --- ## Resources You'll Need | Resource | Role | Label | Price | | --------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | ----------- | --------------------- | | [ChatGPT](https://chatgpt.com/?ref=hyperank.ai) | Conversational image generation and editing | Recommended | Free tier; paid plans | | Midjourney | High-quality stylized images with parameters like --ar and --no | Optional | Paid plans | | [Adobe Firefly](https://firefly.adobe.com/?ref=hyperank.ai) | Image generation inside the Adobe workflow | Optional | Free tier; paid plans | | Google Imagen prompt guide | Official reference for prompt structure and modifiers | Recommended | Free | | [hyperank AI Prompt Generator](https://www.hyperank.ai/tools/ai-visibility/ai-prompt-generator?ref=hyperank.ai) | Generates the text questions buyers ask AI engines about your category (not image prompts) | Optional | Free | A note on fit: hyperank is not an image tool. We built it because we think brands need to see how AI engines describe them. It checks what six major AI engines say about your brand and competitors every day, including what they get wrong. That matters once your content and visuals are out in the world and people start asking AI about you. --- ## Troubleshooting Common Issues ### The image looks generic, like stock photography **Likely cause:** The prompt names a subject but no style, lighting or camera detail. **Fix:** Add one medium, one lighting term and one camera term, for example "editorial photograph, golden hour light, 85mm lens". ### Hands, faces or text come out distorted **Likely cause:** These are known weak spots, especially in busy scenes. **Fix:** Use a closer or simpler composition, add a negative prompt for "distorted hands, extra fingers", and add text later in a design tool. ### The model ignores part of the prompt **Likely cause:** Too many competing details, or the key detail buried at the end. **Fix:** Move the most important detail to the start, cut secondary ones, and generate again. --- ## Conclusion You don't need a special ai image prompt generator to get great results; you need a consistent formula. Describe the subject and action, setting, style and mood, lighting and camera, and composition, then use negative prompts to remove what you don't want. ### Key Takeaways - **Outcome:** A five-part formula and copy-ready templates that work across ChatGPT, Midjourney, Firefly, Imagen and Stable Diffusion. - **Key insight:** Concrete visual words (lens, light, angle, palette) control the result far more than abstract goals. - **Next action:** Take one template from Step 6, fill it in for a real project today, and save the version that works. --- ## FAQ ### How do you write AI image prompts, and are there free templates and examples? How to Write AI Image Prompts (Free Templates and Examples) comes down to one formula: describe the subject and its action, the setting, the style or medium, the lighting and mood, and the camera view and composition, including the aspect ratio. For example: "A matte white ceramic mug on a light oak table, studio product photography, sage green background, soft diffused lighting, centered, 4:5." Add a short negative prompt for anything you want removed, generate a few versions, and change one detail at a time. The templates in Step 6 cover lifestyle photos, product shots, blog illustrations, flat lays and concept art. ### How do AI image prompts work? Image models learn from large collections of images paired with text descriptions. When you write a prompt, the model creates a new image that matches the visual patterns associated with your words, which is why specific, visual words work better than abstract ones. ### What is the best prompt generator for AI art? Many people use ChatGPT or another chat assistant as a prompt generator: describe your idea and ask it to write a detailed prompt using the subject, style, lighting and composition formula. The templates in this guide do the same job for free and keep your style consistent. ### Are AI prompters artists? Opinions differ. Writing good prompts takes skill in visual vocabulary, composition and iteration, but most people see it as directing a tool rather than making art by hand. Copyright rules for AI-generated images are also still developing, so check current guidance before using them commercially. ### How long should an AI image prompt be? One to three sentences is enough for most images. Google notes that Imagen works with both short and long prompts; start short and add detail where the result misses. ### What are the best prompts for social media images? Start with the platform's aspect ratio (4:5 for Instagram feed, 9:16 for Stories, 1.91:1 for LinkedIn), keep one large, simple subject, use your brand palette and leave space for text. --- *Methodology: This guide draws on Google's official Imagen prompt guide, documented parameters in Midjourney and Stable Diffusion, and usage figures reported by TechCrunch. Tool features and free tiers change often, so check each provider's current documentation and terms before relying on them. We have not benchmarked specific models against each other.*