MidJourney is an artificial intelligence tool that generates images from text. However, unlike Dall-E, this AI has not been trained to aim for realism, but rather for artistic beauty. The result is stunning, and many artists are impressed. Find out all you need to know.
Since June 2022, the artificial intelligence program Dall-E Mini aka Craiyon has been a viral success on social networks. This AI is capable of create images from text and has given rise to many hilarious memes.
A few weeks after Dall-E Mini, a new text-to-image generator based on artificial intelligence, is creating a buzz on the Internet. It’s called MidJourney.
What is MidJourney?
MidJourney is an artificial intelligence system capable of create images from text entered by the user. It represents ” an independent research laboratory exploring new ways of thinking. Extending the imaginative powers of the human species “.
This artificial intelligence model has been trained on a large number of imageslike most text-to-image AIs. However, it distinguishes itself by focusing on artistic style rather than realism. In optimizing their AI, MidJourney’s creators wanted above all to to generate beautiful images.
Currently in version 3.0, the software now incorporates a loop of feedback loop based on activity and reactions of users. This has helped to improve image quality, by analyzing data indicating which images users like and how they use them.
The company has not yet revealed which technologies it uses. However, it confirms that it uses most advanced artificial intelligence models with billions of parameters, trained on billions of images. What’s more, the images are generated on a vendor’s cloud using green energy.
Each image requires several petaops, which represents 10^15 operations per second. According to the software’s creators, never any service accessible to the general public has ever enabled an individual to use so much computing power.
What is MidJourney for?
Most Internet users use this tool to have fun, and bring their imaginations to life. However, about 30% of users use it for professional purposes.
Many graphic designers use MidJourney as part of their concept development workflow. They generate several variations of an idea, and present them to their clients to determine which direction to pursue. According to Holz, professionals use this tool to ” supercharger “the creation process or communication.
In addition, according to the creator, approximately 20% of users would use it for therapeutic purposes. Creating images could help them overcome trauma or grief. AI is used here as a ” a tool for emotional and intellectual reflection “.
For a long time, MidJourney was reserved for non-commercial use.. Now, commercial use can be authorized. To this end, the well-known magazine The Economist let AI design its cover in June 2022.
Midjourney price plans
As of April 2023, the free trial version will no longer be available. You will need to register for a package from $10 per month. It allows users to create up to 200 images per month. After 200 images, users can continue generating images for $4 per 60 images. It also offers free passes for friends and access to the members’ gallery.
Submitted to 30 per monthThe standard subscription offers unlimited personal users, free passes for friends and access to the members’ gallery. It also includes 900 minutes of accelerated time, enabling users to generate images as quickly as possible.
Plan Pro was launched in mid-December 2022, at 60 $ per month. It offers fast GPU hours: 30 hours/month and unlimited hours of relaxed GPU. Turbo Plan launched in July 2023 costs 120 $ per month. It offers 60 hours of fast GPU use and unlimited hours of relaxed GPU use.
Midjourney also offers annual plans with a 20% discount. The basic annual subscription costs 8 $ per monthagainst 24 $ per month for the standard offer. The annual subscription for the Pro package is 48 $ per month.
What content is prohibited?
As a social application, MidJourney sets rules and limits regarding permitted content. Visit gore or adult content is strictly forbidden. In addition, users are asked to avoid creating visually shocking or disturbing content.
Certain texts may be blocked automatically. In addition, some forty moderators keep an eye on images created by users.
The creators of MidJourney also resent the idea of their AI being used to create fake photographs type DeepFake. According to them, this is a usage “ extremely dangerous “.
Who owns the intellectual property of the art created by MidJourney?
The intellectual property is a sensitive issue when it comes to AI-generated art. At present, for example, the US justice system prohibits the granting of copyright on AI-generated images.
In February 2022, the US Copyright Office Review Board rejected a request to grant a copyright to a computer-generated landscape image entitled ” A Recent Entrance to Paradise “. For good reason: this image had not been created by a human.
According to MidJourney terms of use, users own all works they create with the service. However, the company requires a copyright license from users to reproduce content created with the service. This is a necessary precaution for hosting users’ images.
In addition, intellectual property issues may arise in relation to AI models. trained on copyrighted material. AI trained on existing content also runs the risk of generating images with similarities…
The creators of MidJourney acknowledge that the law is not yet clear enough regarding AI-generated content. He expects more specific legislation to be passed in the future.
David Holz: who is the creator of MidJourney?
David Holz grew up in Florida. He started his design company while studying mathematics and physics at university. He was working on his PhD in Mathematics, but eventually took a break. in 2008 to co-found Leap Motion. The latter was dedicated to the manufacture of computer peripherals.
The following year, Holz spent a year as aresearch student at the Max Planck Institute. He then spent two years as a research student at NASA’s Langley Research Center. He also worked on LiDAR, missions to Mars and atmospheric science.
However, this was also a time of doubt for Holz. He explains that he wondered “why he was working on these things”. Eventually, he wanted to work on just one thing that was close to his heart.
So he focused on Leap Motion. This company is developing a hardware device that can track hand movements and use them to control a computer or other device.
Holz ran the company for twelve years. When he left, it employed around 100 people. In 2021, he finally to create MidJourney.
According to him, the company is still relatively small, with only a dozen employees. It is also self-financed, and has no investors.
In any case, Holz explains that the team’s motivations are not really financial. He says “ we’re just here to work on things we’re passionate about and have fun. And we’re working on a lot of different projects “.
MidJourney vs DALL-E
DALL-E (L) vs Midjourney (R) ๐งต
MJ has a certain “je ne sais quoi”, the imperfections are more beautiful, a bit like an analog synth. It’s often more contextually creative, and amazing w textures / vibe
DALL-E deals better with very clearly instructed scenes
Same prompt:
– fabians.eth (@fabianstelzer) July 5, 2022
While Dall-E focuses on realism of the images generated, MidJourney places more emphasis on the artistic dimension. This AI can adopt real artistic styles to create an image from a combination of all the desired elements.
To demonstrate the difference between the two, a user named Fabian Stelzer entered the same text on both programs.
According to him, “MidJourney has a certain je ne sais quoiimperfections are more beautiful, a bit like on an analog synthesizer. It’s often more contextually creative and great with textures. DALL-E is better for scenes with very clear instructions.
Texts submitted to both AIs for this test included “Mozart playing at Top of the Pops, 1993” and “a complex installation made of plastic bags and mirror fragments, painted in neon color, studio lighting.”
For the plastic bag installation, MidJourney succeeded in create a much more realistic image. The studio lighting is particularly well done.
The AI-generated art phenomenon
This image-creation AI craze began in 2021 with the launch of CLIP (Contrastive Language Image Pre-Training) by OpenAI. This program was designed to evaluate how well generated images aligned with text descriptions, but then artist Ryan Murdock that the process could be reversed an AI can produce an image from a text.
The generative art community then began a period of unbridled explorationPython code for creating images using a variety of models and techniques.
According to David Holz, creator of Midjourney, it was at this time that it was realized that ” some areas of AI were progressing in really interesting ways, and one of them was the ability of AI to understand language “.
In particular, he refers to transformers: a type of Deep Learning model on which CLIP is based. Similarly, diffusion models are an alternative to antagonistic neural networks (ANNs).
This expert has been blown away by CLIP-guided broadcastingdeveloped by Katherine Crawson, known on Twitter as @RiversHaveWings.
The future of MidJourney
As for the future, MidJourney’s creators fear that it will fall victim to its own success. Right now, several hundred thousand people already use the service and require the power of around 10,000 servers.
However, if 10 million people try to use this technology, Holz explains that there will simply be no not enough computers available in the world.
According to Holz, AI tools such as MidJourney can help artists improve in their discipline. They are not necessarily intended to make everyone a professional artist, and an artist using these tools will always be better than a private individual using them.
On the other hand, he recognizes that these tools could put additional pressure on artistseven if this is not yet the case. He believes that the tool will improve enormously over the next two years…
MidJourney’s best creations
By creating a MidJourney account, you can view other users’ creations on the Community Feed page. New AI-generated images are constantly pouring in. Here’s a selection of impressive creations or suprising made by MidJourney and relayed on social networks…
The infinite shades of disgust ๐คข#midjourney pic.twitter.com/DNV9PzzB8o
– Frederik Vanhoutte – Winterbloed (@wblut) March 21, 2022
#midjourney hallucinating my work in a photorealistic style. pretty cool tech right here ๐คฉ pic.twitter.com/BPBYKFuzpP
– VES3L (@VES3L) March 21, 2022
MidJourney V4: all new features
From Saturday, November 5, 2022, MidJourney has begun alpha testing of its V4. This new version offers more detailed results, and makes it very easy to obtain high-quality results from simple texts.
This major update succeeds V3, available since August. Several thousand members of the official Discord server will be able to test this fourth version, simply by adding “-v 4” to their prompts.
According to founder David Holz, “ V4 is an entirely new code base and a completely new AI architecture. This is our first model trained on a new Midjourney AI supercluster, which we’ve been working on for 9 months. “.
The images produced by this V4 model are much more detailed. Prompts seem better constructed, scene composition improved, and proportionality also appears more realistic.
Another new feature is greatly increased knowledge in terms of locations, creators and much more. Small details are much better represented, and complex promptings with multiple levels of detail are better supported.
This V4 is also better suited to scenes with multiple objects or peopleand offer advanced features such as multi-prompt or image prompting.
In general, this update is warmly welcomed by users. This is an alpha, and the new model will continue to be improved over the coming weeks. In particular, the firm plans to increase the definition and quality of upscaled images, add custom aspect ratios as on the V3, increase image precision, and reduce text artifacts.
MidJourney bans words related to the reproductive system to avoid porn content
The pornographic content generation has always been forbidden on MidJourney. However, users are clever and have been quick to find ways around this restriction.
At the end of February 2023, the platform finally decided to ban words related to the human reproductive system to combat this abuse.
From now on, terms such as “placenta”, “fallopian tubemammary glands”, “sperm”, “uterus”, “cervix”, “hymen” and “vulva” can no longer be used as prompts. In some cases, users attempting to use any of these words will be blocked for a limited time.
According to David Holz, this is a temporary measure to prevent the creation of violent or sexualized images while the company improves AI. The list of prohibited words can be consulted at the dedicated page to MidJourney’s terms of use.
Cyberpunk: Peach John, the first manga created with MidJourney
The author of “ Cyberpunk: Peach John “nicknamed Rootportadmits he has no talent for drawing. To create his comic strip, he turned to MidJourney and other AI tools. These included Stable Diffusion and DALL-E 2. In just six weeks, he was able to create a manga of over 100 pages.
Hardly minute, Rootport had already created the hero of his story using prompts containing keywords such as “pink hair”, “Asian boy” or “stadium jacket”. He then assembled the best images into a comic-book format to produce a book.
According to the author, generative AIs could enable many people to without artistic talent to enter the manga industry as long as they have good stories to tell. In his words, this creative process is akin to a lottery or the incantation of magic spells. It allows him to generate images from his imagination. He admits, however, that the satisfaction of drawing oneself is undoubtedly superior.
Of course, this first AI-drawn manga raises concerns for this multi-billion dollar industry. Professor Satoshi Kurihara told AFP that mangaka’s junior assistants risk being replaced by AI…
MidJourney upgrades to version 5: photorealism and (almost) normal human hands
March 2023, MidJourney announced the launch of version 5 of its text-to-image service. This new version can produce photo-realistic images, so much so that some users find them almost “too perfect”.
We note a clear improvement compared with the first version launched in March 2022, but also with version 3 launched in August and version 4 in November. Each iteration adds more detail to the results generated.
Visit skin textures and facial characteristics are very realistic, as is the lighting. The reflections and shadows are better, and viewing angles are more successful.
In addition, the eyes are now almost perfect and hands are finally normal (or almost). This is a major step forward, since it has been one of the tool’s main weaknesses until now.
Visit hands now have 5 fingers most of the time, instead of 7 or 10 as was often the case before.
MidJourney also points out that v5 offers a much wider range of styles and is more sensitive to prompting. It also generates less unwanted text and offers twice the image definition.
MidJourney v5 is available as alpha test for subscribers to the service. It can be accessed from Discord.
MidJourney launches a magazine created entirely by AI
Following consultation with its community, MidJourney has decided to launch a monthly print magazine of the same name.
Each month, readers can discover a selection of images created with the help of AI, highly rated by users. The publication will also include sample prompts and interviews with the image creators.
L’subscription will cost $4 per monthbut early subscribers will receive the first issue free of charge.
MidJourney pays for itself, the fault of cheaters
March 30, 2023, MidJourney has decided to end free trials of its service. CEO David Holz cites ” extraordinary demand and abuse of trials “.
According to him, this pause is due to ” a massive number of people creating disposable accounts to get free images “.
He also points out that the culprit could be ” tutorial video goes viral in China “. However, this wave of abuse occurred at the same time as a ” temporary GPU shortage “.
The combination of these two hazards resulted in a service outage for paying users. This explains why the company has decided to discontinue free trials altogether.
The company points out that it is still looking for a way to offer free trials again. In particular, it tried to require an active email address, but this was not enough, so it will be necessary to find a viable alternative…
MidJourney bans images of Xi Jinping and creates controversy
While MidJourney willingly allows DeepFakes of the Pope, Macron or Donald Trump, the freedom is far from total on the platform.
Anyone attempting to generate an image from a prompt containing Xi Jinping’s name will receive a warning message.
Many users deplore a form of censorshipin opposition to the principles of freedom of expression.
Back in June 2022, CEO David Holz explained that he simply wanted to minimize scandals and drama.
At the time, he highlighted the difference in the way in which political satire in the West and in China. What may seem harmless in our country can pose a problem over there, and even endanger users…
The new “/Describe” feature converts an image into a prompt
After text-to-image, MidJourney has now reversed its formula with a new image-to-text feature launched in early April 2023. Visit new “/describe” command allows users to harness the power of artificial intelligence to transform images into words.
This new tool has many advantages for a wide variety of use cases. First of all, it could improve web accessibilityby enabling the visually impaired to better understand what is displayed on their screens.
Until now, their only option was to consult the “alternative text associated with images to describe them. However, this description had to be done manually by web designers.
Midjourney just launched a new command:
/describe.
It’s basically a super powerful tool for reverse-engineering any image.
Use it at your own discretion… ๐ pic.twitter.com/InHmViyilB
– Pietro Schirano (@skirano) April 4, 2023
MidJourney will enable automate the task to save precious time. Beyond accessibility, search engines will be able to index images more efficiently based on these complete AI-generated descriptions.
Likewise, image captions can be used to explain images and provide more context for visitors. This feature could therefore greatly enhance the web as a whole,
In addition, image-to-text generation creates a loop of feedback loop for the text-to-image system from MidJourney. Users will be able to use these descriptions to write prompts to create variants of existing images…
As is the case for images, MidJourney creates four descriptions for each photo uploaded. It is also possible to generate new variants of one of them, or edit it manually.
To load an image simply type “/describe in the text field. A drag-and-drop download field appears.
Midjourney 6, the latest version of the popular image generator
The version 6 of Midjourney was released on December 20, 2023 in alpha version. Some users are already delighted with the improvements it brings.
These new features include significantly improved, more realistic and highly detailed images. In addition, the can generate readable text in imageswhich has eluded Midjourney since its release in 2022.
The update does not take effect by default. You need to enter the command “/settings” on the Midjourney Discord server or in a direct message (DM) to the Midjourney bot, then use the drop-down menu at the top to select V6. You can also proceed the old-fashioned way by manually typing “-v 6” after the prompts.
Our blog is powered by readers. When you purchase via links on our site, we may earn an affiliate commission.