How to write Midjourney prompts for V8.2: prompt structure, key parameters, 18 marketing examples, commercial use rules and image SEO tips.

In short
a good Midjourney prompt is a short, specific description of the subject, the medium, the setting, the light, the colors, the mood and the framing, followed by parameters such as --ar for the format, --s for the level of styling and --raw for a more literal, photographic result. Since July 24, 2026 the default model is V8.2, which also brings the Edit Model for changing an image with written instructions. You own the images you create, but a company with more than $1,000,000 in annual revenue needs a Pro or Mega plan to use them commercially.
Want to go further? Ask
Midjourney prompts are the text instructions you type in the Imagine bar on midjourney.com or in Discord to generate an image or a video. Midjourney's own prompt basics guide gives the core rule: short, clear phrases beat long lists of instructions. This guide applies it to marketing visuals, with every version and parameter checked in the official documentation in October 2026. If you also use Google's model, compare with our Nano Banana prompts, which favor full sentences instead.
Midjourney releases new models regularly and keeps older ones available through the --v parameter or the settings panel. The versions you will meet:
| Version | Released | Status (October 2026) | What it is good at |
|---|---|---|---|
| V8.2 | Default since July 24, 2026 | Default | Bolder, more sophisticated aesthetics, better Personalization, new Edit Model |
| V8.1 | April 14, 2026 | Available | Fastest model (standard jobs about 4 to 5 times faster than earlier versions), strong prompt adherence, native 2K HD images |
| V7 | April 3, 2025 | Available (--v 7) | Draft Mode at half the GPU cost, Omni Reference |
| V6.1 | July 30, 2024 | Available | Coherent details, about 25% faster than V6 |
| Niji 7 | January 9, 2026 | Available (--niji 7) | Anime and illustrative styles, more literal prompt reading |
Source: Midjourney's Version documentation, checked October 2026. For marketing work, stay on the default V8.2 unless you need a feature only an older version offers, such as Draft Mode in V7.
Midjourney's documentation lists seven areas to describe when a detail matters to you. Use them as a checklist, not as a template to fill in every time:
Three rules from the official guide make the biggest difference. First, describe what you want rather than what you do not want: writing "no cake" can still produce a cake, so use the --no parameter to exclude elements. Second, choose precise words ("enormous" rather than "big"). Third, put parameters at the very end, after a space, without commas or periods. For text inside the image, write the words between double quotation marks, keep them short, and add a cue such as "with the words". This is the same discipline as prompt engineering for text models: one clear intent per prompt, then iterate.
Parameters change how the image is generated. The full list is in the official parameter list; these are the ones marketers use daily:
| Parameter | What it does | Values and default |
|---|---|---|
| --ar (or --aspect) | Sets the format of the image | Default 1:1; for example 16:9 for a blog header, 4:5 for an Instagram post, 9:16 for a story |
| --s (or --stylize) | How much artistic interpretation Midjourney adds | 0 to 1000, default 100; lower values follow the prompt more literally |
| --raw | Turns off the default Midjourney styling | On or off; more realistic, photo-like results with simple prompts |
| --c (or --chaos) | Variety between the four images of a grid | 0 to 100, default 0 |
| --no | Excludes elements | A list separated by commas, such as --no text, people |
| --v (or --version) | Chooses the model | Default V8.2; --v 7 to use V7 |
| --sref | Copies the style of a reference image (colors, medium, textures, light), not its objects | An image URL; --sw controls its strength |
| --p (or --profile) | Applies your Personalization profile or a moodboard | Your profile, built by rating images |
| --hd / --sd | Image resolution | HD 2048 px (1.3 GPU minutes) or SD 1024 px (0.8 GPU minutes) |
| --edit | Edits an image with written instructions or combines up to four reference images (V8.1 and V8.2) | One to four image URLs in Discord; attach images on the web |
| --tile | Creates a seamless repeating pattern | On or off; not compatible with the Edit Model |
| --seed | Reuses the same starting noise to test small prompt changes | A number |
The maximum aspect ratio is 14:1 (4:1 in HD), according to the version compatibility chart.
Each prompt below works as is on V8.2. Replace the subject, colors and format with your own, then iterate from the best of the four images. For more ready-made prompts across text and image models, see our SEO and GEO prompt library and our selection of AI prompts.
For the captions and formats that go with these visuals, see our social media content ideas.
Midjourney is one tool among many: our comparisons of the best graphic design software and of AI content creation methods help you pick the right one for each format.
Yes, with conditions set by the Midjourney Terms of Service (version effective May 27, 2026) and its help article on using images and videos commercially:
| Situation | Rule |
|---|---|
| Individual or company with up to $1,000,000 in annual revenue | You own your images and can use them commercially on any plan, even after cancelling |
| Company (or its employee) with more than $1,000,000 in annual revenue | A Pro or Mega plan is required to own the images |
| Upscale of another user's image | The image stays the property of its creator: you need their permission |
| Visibility of your images | Public and remixable by default; Stealth Mode, available on Pro and Mega only, keeps them private on the website |
| License granted to Midjourney | Midjourney receives a perpetual license to reproduce and display your prompts and images |
Plans cost $10 (Basic), $30 (Standard), $60 (Pro) and $120 (Mega) per month, with a 20% discount on annual billing (Midjourney plan comparison, checked October 2026). Ownership under the terms is not the same as copyright protection: in Part 2 of its report on copyright and AI (January 29, 2025), the US Copyright Office concluded that AI outputs are protected only where a human author has determined sufficient expressive elements, which excludes the mere provision of prompts. Laws differ by country, and Midjourney itself recommends consulting a lawyer. Never prompt a competitor's logo, a protected character or a real person's likeness for an ad.
A generated image only helps your rankings once it is published correctly. Google's image SEO best practices apply to AI images exactly as to photos:
AI assistants are increasingly multimodal, but they cite pages for their text: the image supports the article, and the article earns the citation. Our image SEO guide covers the rest, from image sitemaps to lazy loading.
Start with a short prompt that covers subject, medium, light and framing, add --ar for the format, then adjust --s, --raw and --sref until the result matches your brand. The best Midjourney prompts for marketing are short, specific and tied to one format, with parameters doing the technical work. Images make an article more attractive; the article itself is what ranks on Google and gets cited by ChatGPT or Gemini. To publish those articles at a steady pace, the Sorank article generator writes and publishes sourced, illustrated posts in autopilot, and Sorank also tracks whether AI assistants cite your brand.
Midjourney's documentation recommends short, simple prompts made of clear phrases rather than long lists of instructions. A one-word prompt works and lets the default style fill the gaps, while each detail you add (subject, medium, light, colors, framing) gives you more control and less variety. For marketing visuals, one or two lines plus a few parameters is usually enough.
Yes, since version 6: write the words between double quotation marks, for example a poster with the words "Summer Sale". Rendering works best with short words in the Latin alphabet, and Midjourney suggests using Raw or a lower Stylize value if the text comes out wrong. For long copy, add the text afterwards in a design tool.
Stylize (--s, from 0 to 1000, default 100) sets how much artistic interpretation Midjourney adds: low values follow the prompt more literally, high values give more creative results. Raw (--raw) turns off Midjourney's default styling altogether, which gives more realistic, photo-like images with simple prompts. Marketers often combine --raw with a low stylize value for product and lifestyle photos.
THEY TRIED SORANK BEFORE YOU
