The short answer
Midjourney is an AI image and video generator from Midjourney, Inc., used on midjourney.com or in Discord. Its default model since 24 July 2026 is V8.2, which focuses on aesthetics, image quality and Personalization, and an edit model added in August 2026 changes images from plain instructions. There is no free plan or web trial, and images are public unless you use Stealth Mode on the Pro or Mega plan.
| Made by | Midjourney, Inc., United States |
|---|---|
| Free plan | No. A limited trial exists only in the niji·journey mobile app |
| Models | V8.2 (default since 24 July 2026), V8.1, the V8.2 edit model, Niji 7, V1 video model |
| Best at | Art-directed images, concepts and mood boards, consistent styles |
| Shows sources | No, it generates images rather than answering questions |
| Team plans | No separate team tier; companies over USD 1,000,000 revenue need Pro or Mega |
| Your data | Images are public on midjourney.com unless you use Stealth Mode (Pro and Mega only) |
What Midjourney is in late 2026
Midjourney is an AI image and video generator made by Midjourney, Inc. You type a description, and it returns a grid of images, usually with a look that is more art-directed and more striking than most rivals manage. You can use it on midjourney.com, which is now the main way to create, or through the Midjourney Bot in Discord. Your settings are shared between the two.
The model has moved fast this year. Midjourney released the first V8 alpha on 17 March 2026, followed by V8.1 on 14 April, which became the default on 11 June, replacing V7. On 24 July 2026 Midjourney launched V8.2, which focuses on aesthetics, image quality and Personalization, and it is now the default model. Midjourney says V8.2 images should be "more creative, bold, sophisticated, edgy and fresh" and that random low-quality results should be far less common.
On 27 August 2026 Midjourney opened its V8.2 edit model to everyone. It edits images from plain instructions, generates from up to four reference images at once (this replaces the old Omni Reference), and handles inpainting and outpainting. Video comes from the V1 video model: you animate an image into a five-second clip and extend it.
What Midjourney is best at
Images with a strong, deliberate look
Midjourney's strength is taste. For mood boards, campaign concepts, editorial illustrations, book covers, album art, interior and product concepts, it often produces something usable in the first grid. Midjourney's own launch notes say V8 does best with longer, more specific prompts and with its style systems, so describe the subject, the setting, the light, the lens or medium and the mood.
A consistent style across many images
This is where professionals get the most value. Three tools keep a look consistent. Personalization learns your taste from images you rate, and V8.2 "understands" it much better if you have rated a lot. Moodboards let you upload reference images that define a style. Style references, added with --sref and a code or image URL, reuse a style across prompts, and --sref random gives you a random style code to explore. For a brand team, a saved style code is a quick way to make twenty social images that look like they belong together.
Editing instead of starting again
With the edit model, you can open an image and type what to change, such as "make it dusk and add rain on the window". Since September 2026 targeted edits change only the pixels you select, so you can make several edits without the image degrading. On Discord, the equivalent is --edit followed by an image URL.
Text, patterns and short video
Text rendering improved with V8: put the exact words in quotation marks. --tile makes repeating patterns, and Midjourney says tiles now join without visible seams on V8.1 and V8.2. For video, clips start at five seconds and can be extended four times, four seconds at a time, up to 21 seconds. Add --motion low or --motion high, --loop for a looping clip, or --end to set an end frame.
Where Midjourney falls short
- No free way to try it on the web. Midjourney's docs say there is no free trial on midjourney.com or Discord. A limited trial exists only in the niji·journey app for iOS and Android.
- Your images are public by default. Images appear on Explore and your profile unless you use Stealth Mode, which only Pro and Mega include. Midjourney's own October changelog warns that putting images in a private folder does not change this.
- Precision is not its job. If you need an exact layout, a correct logo or a product shot that matches reality, a design tool or a photo shoot is still the better route. Midjourney interprets; it does not follow a spec.
- Text is better, not perfect. Short headlines in quotation marks usually work. Paragraphs, small print and charts do not belong in Midjourney.
- Plan limits shape how you work. Relax mode, where jobs wait in a queue instead of using your Fast GPU time, is not on the Basic plan, Relax video needs Pro or Mega, and HD (720p) video needs Standard or higher and only runs in Fast mode.
How to write better Midjourney prompts
- Describe, do not command. "A ceramicist's studio at golden hour, dust in the light, shot on 35mm film" beats "make a nice studio picture". Midjourney rewards concrete nouns and visual details.
- Put parameters at the end. Parameters start with two hyphens and go after your description. --ar 16:9 sets the aspect ratio (the default is 1:1, and decimals are not allowed).
- Control how much Midjourney adds. --stylize (or --s) runs from 0 to 1000 with a default of 100. Lower values follow your words more literally; higher values add more of Midjourney's own flair. --raw strips back the default look, which helps for photorealism and plain product images.
- Explore first, then tighten. --chaos from 0 to 100 makes the four images in a grid more different from each other. Use around 30 when you are exploring, then drop it to 0 once you have a direction. --weird, from 0 to 3000, pushes into unusual territory and is experimental.
- Exclude things explicitly. --no followed by a comma-separated list, such as --no text, people, removes unwanted elements.
- Save your defaults. On the web you can go to Settings, then Advanced, then Your defaults, or click "Set current pills as default", so your usual aspect ratio and style are always on.
- Use HD only for finals. --hd renders at 2048px and --sd at 1024px. Midjourney says SD takes about four seconds and HD about twelve on V8.1, so draft in SD and rerun the keeper in HD.
Want a guided start? Our Midjourney lessons build these habits step by step, and best AI for images compares it with the alternatives.
A simple Midjourney workflow for a work project
Say you need six images for a product launch: a hero banner, four social posts and a slide background. This is the order that wastes the least time and GPU budget.
- Explore in standard resolution. Write one descriptive prompt and add --chaos 30. Run it two or three times and save the grids you like to a folder.
- Lock the look. When one image has the right feel, use it as a style reference, or try --sref random until a code fits and write that code down. If you have rated plenty of images, switch Personalization on as well.
- Set your defaults. Save the style code and your usual settings as defaults so every new prompt starts from the same place.
- Make the set. Change only the subject and the aspect ratio for each format: --ar 16:9 for the banner, --ar 4:5 for feed posts and --ar 9:16 for stories.
- Fix, do not reroll. Use the edit model for small problems, such as an extra object or the wrong colour, instead of starting again and losing the composition.
- Finish in HD. Rerun only the final picks with --hd, then add your logo and copy in a design tool, where the layout is exact.
Midjourney privacy settings and commercial use
Privacy in Midjourney is mostly about visibility. On the Basic and Standard plans, everything you make can appear in the public gallery. On Pro and Mega, you can turn on Stealth Mode in the settings panel from the settings button in the Imagine bar on the web, type /stealth in Discord, or add --stealth to a single prompt. Stealth does not hide images you already made in public, so you need to make those private by hand. It also does not hide images generated in shared Discord channels, which other members can see. For full privacy on Discord, use direct messages with the Midjourney Bot or a private server.
For business use, the rule to know comes from Midjourney's Terms of Service, dated 27 May 2026: a company, or an employee of a company, with more than USD 1,000,000 in gross revenue a year must be on the Pro or Mega plan to own the images it creates. If you work for a business of that size, a Basic subscription is not enough, however light your use. Never upload confidential product designs or client material on a plan without Stealth Mode. Even with Stealth Mode, the same Terms give Midjourney a perpetual, worldwide licence to the content you put in and the images you make, and Midjourney's public summary of its training content lists prompts and preference data among the data its models train on. Stealth Mode controls who can see your images, not what Midjourney may do with them.
Which Midjourney plan suits whom
- Basic: for individuals testing the waters. It has Fast GPU time only and no Relax mode, so heavy weeks run out.
- Standard: for regular personal or freelance use. It adds unlimited image generation in Relax mode, where jobs wait in a queue.
- Pro: the sensible choice for professionals and any company over the USD 1,000,000 revenue line. It adds Stealth Mode and Relax video.
- Mega: for studios and heavy users who need the most Fast time, with Stealth Mode and Relax video.
Who it is not for: if you need a free tool, exact brand layouts or images made inside your office apps, look at our Midjourney alternatives or compare tools side by side. Check Midjourney's plan comparison page for current prices and GPU hours.
What is new and still in testing
Several features are on Midjourney's alpha site, alpha.midjourney.com, rather than the main site, so they may change. Shared folders, launched on 7 October 2026, let you invite people as Collaborators, who can generate into the folder, or Viewers. Thinking Mode lets you rerun a V8.2 job with "thinking": Midjourney looks at its own image, finds where it missed your prompt, such as wrong objects, layout, anatomy or text, and generates again. Live style previews in the Styles sidebar show your prompt across styles before you spend a generation. If something on this page looks different on your screen, the alpha site is usually why.
Prompts to copy
Editorial hero image
A small independent bookshop on a rainy evening, warm light spilling onto wet cobblestones, a cat asleep in the window, cinematic 35mm film photograph, soft depth of field --ar 16:9 --stylize 250
You get: A wide, moody header image for a blog or newsletter
Clean product shot
A matte white ceramic coffee mug on a pale oak table, soft window light from the left, minimal studio product photograph, plenty of empty space on the right for text --ar 4:5 --raw --no text, logo, people
You get: A plain, realistic product image with room for copy
Explore a campaign concept
Concept art for a sustainable sneaker campaign, sneakers made of moss and recycled fabric, set in a city rooftop garden at sunrise, bold colours --ar 1:1 --chaos 30
You get: Four very different directions to choose from
Find a house style
Flat illustration of a team planning a project around a table, friendly and modern, limited colour palette --sref random --ar 3:2
You get: A random style code to keep if you like the look
Poster with a headline
Vintage travel poster of a mountain lake at dawn, bold screen-print style, the words "VISIT LAKE ORRA" in large letters across the top --ar 2:3
You get: A poster with a short headline rendered in the image
Repeating pattern
Hand-drawn botanical pattern of lemons and leaves, cream background, gouache texture --tile --ar 1:1
You get: A repeating tile for packaging, slides or wrapping
Edit an existing image
Attach your image, then type: Change the time of day to dusk, turn on the lights in the windows and add light rain. Keep the building and the composition the same.
You get: A targeted edit that keeps the rest of the image
Short looping video
Animate your chosen image with this prompt: steam rising slowly from the coffee cup, gentle movement only --motion low --loop
You get: A five-second looping clip for social media