Aiodic

Back to the table

89 Creative tool Mj Midjourney Creative Tools

Midjourney

Artistic AI image generation.

What it is

Midjourney is an AI image generator. You type a description, called a prompt, such as "a lighthouse on a cliff at dusk, oil painting, stormy sky", and within seconds it shows you four different pictures that match. You can then ask for variations of the one you like, enlarge it, change part of it, or turn it into a short video clip.

It is made by Midjourney, Inc., a small company based in San Francisco that calls itself an independent research lab. It was founded by David Holz, who earlier co-founded Leap Motion, a company that made hand-tracking sensors. Midjourney opened to the public as a beta test on 12 July 2022. For its first two years the only way to use it was through Discord, a chat app popular with gamers: you joined Midjourney's Discord server and typed a command into a chat channel, where everyone else could watch your pictures appear. In August 2024 Midjourney opened its own website to everyone, and today most people use it there. The Discord route still works.

Midjourney became known for a particular look. Where some image tools aim for plain, literal results, Midjourney's default images tend to be dramatic, richly lit and "finished", like concept art or a magazine photo shoot. That is a deliberate choice by the company, and it is why the tool is loved by some artists and designers and why others find its style easy to spot.

Under the hood it works like other modern image generators. It learned from a very large collection of pictures paired with text, and it creates each new image rather than copying a stored one. Midjourney has not published the full details of its models, but tools of this kind are generally built on the diffusion model approach, which starts from random visual static and cleans it up, step by step, into a picture that matches the words.

Here is what it can do at the time of writing. Features and names change often, so check Midjourney's own documentation for the current list:

Every plan is a paid subscription. At the time of writing there is no free trial on the website or in Discord.

Why it matters

It put professional-looking images within reach of anyone who can describe one. Before tools like Midjourney, a striking illustration meant hiring an artist, buying stock photos, or learning years of craft. Now a small business owner, a teacher or an author can get something eye-catching in minutes. As early as June 2022, before the public beta, the magazine The Economist used Midjourney to create one of its covers.

It changed how creative professionals work, not only who can make pictures. Designers, architects and advertising agencies use it to sketch ideas quickly: dozens of moods for a campaign, rough looks for a building, or visual references to show a client before any real work starts. In 2022 Holz told the technology site The Register that many graphic artists used it to develop concepts, showing clients a few variations to agree a direction before the real work began.

It sits at the centre of the copyright fight over AI. Midjourney learned from images collected from the internet. In a 2022 interview with Forbes, Holz said the company had not asked living artists for consent, and explained that there was no practical way to trace the origin of a hundred million images. Many artists see that as taking their work without permission or payment. Midjourney argues that learning from images is lawful "fair use". Courts are now deciding who is right, and Midjourney is a defendant in several of the biggest cases.

It raises the question of who owns an AI picture. For creators who want to sell what they make, this matters most of all. In the United States, the Copyright Office has repeatedly said that images Midjourney generates from prompts are not the product of human authorship, so they can't be registered for copyright on their own. You may still own them under Midjourney's terms, and you can usually sell them, but you may not be able to stop someone else copying them. The examples below show where the line has been drawn so far.

It made fakes cheap. A believable photo of something that never happened used to take skill. Now it takes a sentence. That has consequences for news, politics and trust, which the second example shows.

Real-world examples

A comic book that was only partly protected (2022 to 2023)

Kris Kashtanova wrote a short comic book called Zarya of the Dawn and illustrated it with Midjourney. In September 2022 Kashtanova applied to register it with the U.S. Copyright Office as its sole author, and the registration was granted. The application did not say the images were AI-made. When the Office learned how the book had been made, it reviewed the registration. On 21 February 2023 it wrote to Kashtanova's lawyer with its decision.

The story text was Kashtanova's own work and was protected. So was the selection and arrangement of the text and images on each page, in the same way that an anthology editor's choice and order of pieces can be protected. But the Midjourney images themselves were not, because the Office concluded that Midjourney users do not control what the tool actually produces. A prompt, it said, works more like a suggestion than an instruction, and the result cannot be predicted in the way an artist controls a brush or a camera. It also said that substantial edits made to an image by hand, those edits could be protected. It cancelled the original registration and issued a narrower one.

For self-publishing authors this is the key ruling. Your words are yours. Your design choices can be yours. The raw AI pictures, at least in the United States at the time of writing, are not protected by copyright on their own. The Office took the same view in 2023 about Jason Allen's prize-winning Midjourney picture "Théâtre D'opéra Spatial", despite the more than 600 prompts he said he used; that story is told on the Diffusion Model page. Allen's court challenge was still undecided at the time of writing. In March 2026 the Supreme Court declined to hear a related case, Thaler v. Perlmutter, leaving in place a lower-court ruling that a work needs a human author.

The Pope in a puffer jacket (March 2023)

In late March 2023, pictures of Pope Francis striding along in a long, puffy white designer-style coat spread across social media. Millions of people saw them, and many believed they were real. They weren't. A 31-year-old construction worker from the Chicago area, who gave his name to BuzzFeed News only as Pablo Xavier, had made them with Midjourney on the afternoon of Friday 24 March and posted them to a Facebook group and to Reddit. He told the outlet he meant no harm and thought the image was funny. It went further than he expected.

The same month, other Midjourney images showing Donald Trump apparently being arrested also circulated widely. On 28 March 2023, Midjourney stopped its free trials. Holz said on Discord that the reason was a combination of extraordinary demand and trial abuse, and he later said the immediate trigger was a wave of throwaway accounts; he did not blame the Pope image. Either way, the episode became a turning point in public awareness: an ordinary person with a few dollars' worth of tools could now produce a news-style photo realistic enough to fool celebrities and journalists. Midjourney has tightened its moderation since, but no filter catches everything.

Disney and Universal sue Midjourney (June 2025 onward)

On 11 June 2025, Disney and NBCUniversal, with several of their studios including Marvel, Lucasfilm and DreamWorks Animation, sued Midjourney in federal court in Los Angeles. It was the first time major Hollywood studios had sued an AI company. Their complaint included pictures of characters such as Darth Vader, Elsa, Shrek, the Minions and members of the Simpson family that they said users had produced with Midjourney, and it called the service "a bottomless pit of plagiarism". The studios argued that Midjourney had copied their films and artwork to train its models and was now selling images of their characters to paying subscribers. In early September 2025, Warner Bros. Discovery filed a similar suit over characters including Superman, Batman, Scooby-Doo and Bugs Bunny.

Midjourney denies infringing. In its August 2025 answer to Disney and Universal it argued that training an AI model on images to learn visual concepts is a lawful fair use, and that the studios' own staff use Midjourney and other AI tools. At the time of writing the case was in the evidence-gathering stage, with arguments over how much the studios must reveal about their own use of AI, and no court had ruled on the main questions. Whatever the result, it will shape what image tools are allowed to produce.

Midjourney is also one of the companies named in Andersen v. Stability AI, the lawsuit brought by a group of artists in January 2023. That case is covered on the Stable Diffusion page.

An everyday scenario: a cover concept for a thriller

This one is illustrative. Imagine Priya, who self-publishes crime thrillers and is planning a new series. She subscribes to Midjourney's cheapest plan for a month. She prompts "a lone figure on a rain-soaked pier at night, harbour lights, cinematic, moody, teal and orange" and gets four images. She likes the third, runs variations on it, then uploads the cover of her previous book as a style reference so the new images share its colour and mood.

She doesn't treat the result as her finished cover. She knows the image will be public on Midjourney's site unless she pays for a higher plan with Stealth Mode, and she knows the picture on its own is probably not protected by copyright. So she sends her three favourite images to her cover designer as a reference, and the designer builds the final cover with licensed photography, custom typography and the series branding. When she uploads the book, she answers the publishing platform's question about AI content truthfully.

Try it yourself

Midjourney is paid, so be honest with yourself about whether a month's subscription is worth it before you start. At the time of writing, the cheapest plan (Basic) costs 10 US dollars a month, there is no free trial on the website, and the only free option is a limited trial in the separate niji·journey mobile app, which specialises in anime-style art. Prices, plans and menus change, so check Midjourney's plans page before paying.

  1. Create an account or log in at midjourney.com. Choose a plan on the subscription page. The monthly Basic plan is enough to learn the tool, and you can cancel before it renews.
  2. Open the Create page and find the Imagine bar at the top. This is where you type prompts.
  3. Start with a plain description. Try: a cosy bookshop on a rainy evening, warm light in the windows, watercolour illustration. Press Enter and watch four images form.
  4. Change one thing at a time. Swap "watercolour illustration" for "35mm film photograph", then for "linocut print". Comparing results teaches you more than any prompt guide. This is iterative refinement in action.
  5. Set the shape. Book covers and phone screens are tall, videos are wide. Add --ar 2:3 for a portrait shape or --ar 16:9 for a wide one, or use the settings panel.
  6. Try a style reference. Upload an image you have the right to use, such as your own photo or drawing, and drop it into the Style Reference slot. Then prompt something unrelated, such as a fox reading a newspaper. Notice how the colours and texture carry over but the content doesn't.
  7. Unlock personalization. On the Personalize page, pick your favourites from the grid of images until the profile unlocks, then turn Personalization on and rerun an earlier prompt. See whether you prefer the results.
  8. Vary and upscale. Pick the best image, try "Vary (Subtle)" and "Vary (Strong)", then upscale your favourite and download it.
  9. Optional: animate it. Choose "Animate" on an image to make a short video clip. Video uses much more of your monthly GPU time (the computing time your plan includes) than a still image, so try one or two.
  10. Read the rules before you use anything commercially. Check the current Terms of Service and your plan, and check what your publishing platform asks about AI content.

Common mistakes and misunderstandings

"I own it, so it's protected." These are two different things. Midjourney's terms say you own the images you create "to the fullest extent possible under applicable law". But in the United States, as the Zarya of the Dawn decision shows, images generated from prompts are not protected by copyright on their own. Owning something the law doesn't protect means you can use and sell it, but you may have little power to stop others copying it. Your own substantial edits, and the way you combine images with text and design, can be protected. Rules differ between countries. For anything valuable, ask a lawyer where you live.

Assuming every plan allows commercial use in the same way. At the time of writing, Midjourney says anyone who has subscribed may use their images in almost any way, including commercially, and keeps ownership after cancelling. The exception is size: a company, or an employee of a company, with more than 1 million US dollars a year in revenue must be on the Pro or Mega plan to own its images. Check the current terms, because they change.

Forgetting that your images are public. By default everything you make on Midjourney can be seen, and remixed, by other users. Only the Pro and Mega plans, at the time of writing, include Stealth Mode, which keeps your images out of public view. Images made in a shared Discord channel are visible to everyone in that channel regardless. If you're working on an unannounced book cover or a client project, this matters.

Forgetting the licence you give Midjourney. The terms also give Midjourney a broad, permanent licence to use your prompts, any images you upload, and what you make. Don't upload anything you aren't entitled to share.

Prompting for famous characters, logos or living artists' styles. Asking for a well-known character is exactly what the studios' lawsuits are about, and publishing such images can land you in trouble, not only Midjourney. The terms forbid using the service to violate others' intellectual property. The same caution applies to real people's faces and to prompts that name a living artist to imitate their work.

Skipping the AI disclosure when self-publishing. At the time of writing, Amazon's Kindle Direct Publishing asks you to say when a book contains AI-generated content, and it counts cover and interior images. The answer is private and does not appear on your book's page. Other platforms have their own rules. Check them each time you publish, because they change.

Expecting perfect text and details. Recent versions render short text far better than early ones, and Midjourney says V8 improved it further, but titles and names can still come out misspelled, and hands, crowds and small objects can still go wrong. Most professionals add lettering in a design program instead of asking the image tool for it. Zoom in on every image before you use it.

Treating it as a replacement for a cover designer. A good cover follows genre conventions, works as a thumbnail, and carries readable typography. Midjourney can give you a strong image and a lot of ideas. The design work around it is still a skill.

Thinking the pictures are neutral. Like other image models, Midjourney reflects patterns in the pictures it learned from. Researchers have found that neutral prompts can return results skewed by gender and skin colour. If you ask for "a doctor" or "a CEO", look at who it shows you, and say what you actually want. The Bias page explains why this happens.

Related elements