Midjourney: What is it? Use, Limitations and Alternatives

Midjourney is an AI image generator that creates images from text descriptions. This is a research lab run by David Holz, co-founder of Leap Motion. Midjourney offers a more dreamlike arty style to your demands and has a more gothic look compared to other AI generators. The tool is currently in open beta and can only be accessed through a Discord bot on their official Discord.

To generate images, users use the /imagine command and enter a prompt, and the bot returns a set of four images. Users can then choose which images they want to scale. Midjourney is also working on a web interface.

Founder David Holz views artists as Midjourney's customers, not competitors. Artists use Midjourney for rapid prototyping of concept art that they present to their clients before starting to work on their own. Since all of Midjourney's lineups may include copyrighted works by artists, some artists have accused Midjourney of devaluing original creative work.

Midjourney's Terms of Service include a DMCA Takedown Policy, which allows artists to request that their works be removed from the set, if they believe copyright infringement is evident. The advertising industry has also embraced AI tools such as Midjourney, DALL-E, and Stable Diffusion, among others, which allow advertisers to create original content and come up with ideas quickly.

Midjourney has been used by various people and companies to create images and artwork, including The Economist and Corriere della Sera. However, Midjourney has come under criticism from some artists who feel that it is taking jobs away from artists and infringing on their copyrights. Midjourney was also the subject of a lawsuit filed by a team of artists for copyright infringement.

To start using Midjourney, users need to log into Discord and head to the Midjourney website to join the beta. Once accepted, users will receive an invite to Discord Midjourney and can begin generating images by typing /imagine followed by the desired prompt.

Midjourney hasn't revealed much information about his background and training, but it's speculated that he uses a system similar to Dall-E 2 and Stable Diffusion, scraping pictures and text from the internet to describe them, in using millions of published images for training.

Table of contents

The process used by Midjourney to generate images from text prompts

Midjourney uses a text-to-image AI model to generate images from text prompts. The Midjourney bot breaks down the words and phrases in a prompt into smaller pieces, called tokens, which can be compared to its training data and then used to generate an image. A well-designed prompt can help create unique and exciting images [0].

To generate an image with Midjourney, users must type a description of what they want the image to look like using the “/imagine” command in the Midjourney Discord channel. The more specific and descriptive the message, the more the AI will be able to produce good results. Midjourney will then create several different versions of the image based on the prompt within a minute. Users can choose to get alternate versions of any of these images, or enlarge any of them to get a larger, higher quality image. Midjourney offers fast and relaxed modes, with fast mode being necessary to achieve maximum magnification and produce more images in less time.

Midjourney's AI model uses diffusion, which involves adding noise to an image and then reversing the process to retrieve the data. This process is repeated endlessly, causing the model to add noise and then remove it again, ultimately creating realistic images by making small variations in the image. Midjourney scoured the internet for images and text to describe them, using millions of published workout images.

Midjourney's AI model is based on stable streaming, which is trained on 2,3 billion pairs of images and text descriptions. By using the right words in the prompt, users can create almost anything that comes to mind. However, some words are prohibited, and Midjourney maintains a list of these words to prevent malicious people from creating prompts. Midjourney's Discord community is available to provide live help and plenty of examples for users.

Using and generating images

To use Midjourney AI for free, you must have a Discord account. If you don't have one, sign up for free on Discord. Next, visit the Midjourney website and choose Join Beta. This will take you to a Discord invite. Accept the Discord invite to Midjourney and choose to Continue on Discord.

Your Discord app will automatically open, and you can select the ship-shaped Midjourney icon from the left menu. In the Midjourney channels, locate the newcomer rooms and select one of them to start. When you're ready, type "/imagine" in the Discord chat for your newcomers room.

This will create a prompt field where you can enter the image description. The more specific you are in your description, the better the AI will be able to produce good results. Be descriptive, and if you're looking for a particular style, include that in your description. Midjourney offers each user 25 tries to play with the AI.

After that, you will need to register as a full member to continue. If you'd rather not spend money, it's a good idea to take some time and think about what you want to create on Midjourney.

If you want, you can type "/help" to get a list of tips to follow. It is essential to know the list of forbidden words before using the Midjourney AI, as failure to comply with the code of conduct will result in a ban.

/imagine the command

The /imagine command is one of the main commands in Midjourney that allows users to generate AI-generated images based on their demands. Here's how it works:

Users type the /imagine command in the Discord chat and add the settings they want to use.
The Midjourney AI algorithm analyzes the prompt and generates an image based on the input.
The generated image is displayed in the Discord chat, and users can provide feedback and refine their messages using the Remix feature.
Users can also use additional settings to adjust the style, version, and other aspects of the generated image.

The /imagine command accepts both image and text prompts. Users can add prompts as images by providing a URL or attachment for the images they wish to generate. Text prompts can include descriptions of the image users want to generate, such as objects, backgrounds, and styles. Users can also add additional parameters to the command to adjust the version of the algorithm they want to use, enable the Remix feature, etc.

Examples of the types of images Midjourney AI can create

Midjourney AI can create a wide range of images in different styles, including but not limited to:

Illustrations for children's books, such as the example of "A Piglet's Adventure".
Realistic portraits of people, animals and objects.
Surreal and abstract works of art that mix different elements and styles.
Landscapes and cityscapes that can evoke different moods and emotions.
Black and white photography with intricate detail and cinematic effects.
Images that illustrate futuristic or sci-fi themes, such as the example of an old woman half made of robotic parts and wearing a gas mask.

It is important to note that the quality and style of the images generated by Midjourney AI may vary depending on the quality of the prompts, the version of the algorithm used and other factors. Users should experiment with different prompts and settings to get the desired results.

Combine images in Midjourney

To combine two or more images in Midjourney, you can follow these steps:

Choose the images you want to combine and upload them to Discord.
Copy the links to the images and add them to your /imagine prompt as image prompts.
Add "-v 4" to your prompt if version 4 is not enabled by default.
Submit the command and wait for the image to be generated.

For example, to combine two images, you can use the following command: /imagine –v 1

You can also add additional information, including objects, background, and general art style, to create an entirely new image with its own style. For example: /imagine , cartoon style, cheerful crowd in the background, Tesla logo on the chest, -non costume -v 1

Midjourney also launched a new feature, the /blend command, which allows up to five images to be merged without having to copy and paste URLs. You can enable the /blend command by including the –blend flag in your prompt.

It is important to note that this function only works with version 4 of the Midjourney algorithm, and combining images does not require additional text, but adding information usually results in better pictures. Best results are usually achieved by experimenting with Art Styles and tweaking images with Remix Mode.

Combine more than two images

Midjourney allows users to blend up to five images using the /blend command. However, if users need to combine more than five images, they can use the /imagine command and paste the public image URLs into a row. To combine more than two images using the /imagine command, users can add prompts to the command. For example, to combine three images, the command would be /imagine –v 1.

Users can add more command prompts to combine more images. It's important to note that adding additional information to the prompt, including objects, background, and general art style, can help create an entirely new image with its own style. Best results are achieved by experimenting with Art Styles and tweaking images with Remix Mode

Command /blend in Midjourney

Midjourney's /blend command allows users to blend up to five images by adding easy-to-use UI elements directly into the Discord interface. Users can drag and drop images into the interface or select them directly from their hard drive. Users can also choose the dimensions of the image they want to see generated. If users use custom suffixes, they can optionally add them to the end of the command, as with any normal /imagine command.

The Midjourney team designed the /blend command to effectively examine the “concepts” and “mood” of users' images and attempt to blend them. This sometimes results in surprisingly alluring images, and in other cases, users end up with horrifying images. However, the /blend command does not support text prompts.

The /blend command has limitations. The most obvious is that users can only add five different image references. Although the /imagine command technically accepts more than five images, the more references users add, the less important each one is. This is a general issue with problem dilution and not a /blend specific issue. The other major limitation is that the Midjourney blend command does not work with text prompts. This can be unfortunate for advanced users who rarely just mix two images. However, for users looking to create mashups, this limitation doesn't matter much.

Improve build time

there are ways to improve or optimize the generation time for image creation by Midjourney AI. Here are some tips that can help you:

Use specific and detailed prompts: Midjourney generates images based on user prompts. The more specific and detailed the prompt, the better the results. It also reduces the time it takes to generate an image, as the AI algorithm has a more accurate idea of what the user wants.
Experiment with different quality settings: The –quality parameter adjusts the quality of the image and the time it takes to generate it. Lower quality settings produce images faster, while higher quality settings may take longer but yield better results. It is important to experiment with different settings to find the right balance between quality and speed.
Use Relax Mode: Standard and Pro plan subscribers can use Relax Mode, which costs nothing to the user's GPU time, but places jobs in a queue based on how often the device is used. system. Wait times for Relax mode are dynamic, but are typically between 0 and 10 minutes per task. Using Relax mode can be a good way to optimize build time, especially for users who generate a large number of images each month.
Buy more Fast hours: Fast mode is the highest priority processing level and uses the monthly GPU time from the user's subscription. Users can purchase more Quick Hours on their Midjourney.com/accounts page, which helps ensure their images are generated quickly and efficiently.
Use Fast Relax: Fast Relax is a new feature in Midjourney that allows users to generate images faster by sacrificing some quality. The Fast Relax mode generates images with a quality of around 60%, which can be a good compromise for users who want to generate images quickly but do not want to sacrifice too much quality.

In summary, there are several ways to improve or optimize the build time for creating Midjourney AI images, including using specific prompts, experimenting with different quality settings, using Relax mode, or buying more fast hours, and using the Fast Relax mode.

How accurate are the images generated by Midjourney's AI model?

The accuracy of the images generated by Midjourney's AI model may vary depending on the prompt and the quality of the training data. Users can improve the accuracy of generated images by being specific and detailed in their queries. The more specific and descriptive the prompt, the better the AI will be able to produce good results. Midjourney's AI model was trained on millions of images and text descriptions retrieved from the internet, which may also affect the accuracy of generated images.

It should be noted that the AI-generated images of Midjourney have been the subject of controversy regarding copyright infringement and artistic originality. Some artists have accused Midjourney of devaluing original creative work, while others see it as a tool for rapid prototyping concept art to show to clients before they start working on themselves.

How does Midjourney address concerns about copyright infringement and the originality of AI-generated images?

Midjourney: Copyright infringement and originality of AI-generated images

Midjourney has taken steps to address concerns about copyright infringement and the originality of AI-generated images. Midjourney carefully checks every prompt and every image to make sure there are no copyright issues, using only licensed or public domain content, and doing additional research or by asking the authorization of the rightful owner in case of uncertainty.

Midjourney also encourages the responsibility of its users by urging them to respect copyright laws and to use only images and prompts that they have the right to use. If a user questions the source of a post or image, the platform takes prompt action to investigate and remove any infringing content, in accordance with the Digital Millennium Copyright Act (DMCA) of 1998.

The DMCA provides protective provisions for online service providers, such as Midjourney, who act in good faith to remove infringing content when notified by a copyright holder. Midjourney also has a DMCA Takedown Policy which allows artists to request that their work be removed from the set if they believe copyright infringement is obvious. [2][4].

Midjourney's approach to avoiding infringement is consistent with Supreme Court cases such as Feist Publications, Inc. v. Rural Telephone Service Co., Inc. (1991), where the Court held that originality, not novelty, is the essential requirement for copyright protection, and Oracle America, Inc. v. Google LLC (2018), where the Court held that copying an original work, even for a different purpose, can still be considered copyright infringement.

Midjourney's AI-generated imagery has been the subject of controversy over copyright infringement and artistic originality. Some artists have accused Midjourney of devaluing original creative work, while others see it as a tool for rapid prototyping concept art to show to clients before they start working on themselves. Midjourney's Terms of Service include a DMCA Takedown Policy, which allows artists to request that their work be removed from the set if they believe there is copyright infringement.

How does Midjourney ensure that all licensed or public domain content used to create AI-generated images is properly attributed?

It is unclear how Midjourney ensures that all licensed or public domain content used to create the AI-generated images is properly attributed. However, Midjourney carefully checks each post and image to ensure there are no copyright issues, using only licensed or public domain content, and conducting additional research. or by asking the authorization of the rightful owner in case of uncertainty.

Midjourney also encourages the responsibility of its users by urging them to respect copyright laws and to use only images and prompts that they have the right to use. If a user questions the source of a message or image, the platform takes prompt action to investigate and remove any infringing content, in accordance with the Digital Millennium Copyright Act (DMCA) of 1998.

Midjourney also has a DMCA Takedown Policy, which allows artists to request that their work be removed from the series if they believe there is clear copyright infringement.

The rules that users must respect on Midjourney

Midjourney has established a set of rules that users must follow to ensure a welcoming and inclusive community for all. These rules are as follows: [0][1][2] :

Be kind and respect others and staff. Do not create images or use text prompts that are inherently disrespectful, aggressive, or otherwise abusive. Violence or harassment of any kind will not be tolerated.
No adult content or bloody scenes. Please avoid visually offensive or disturbing content. Some text entries are blocked automatically.
Do not publicly reproduce the creations of other people without their permission.
Pay attention to sharing. You can share your creations outside of the Midjourney community, but consider how others might view your content.
Any violation of these rules may result in exclusion from the service.
These rules apply to all content, including images made in private servers, in private mode and in direct messages with the Midjourney Bot.

Midjourney also has a list of banned words that are not allowed in messages. The list of banned words includes words directly or indirectly related to violence, harassment, gore, adult content, drugs or hate speech. Further, it does not allow prompts that include or are related to aggression and violence.

If a word is on the banned word list or if it is closely or remotely related to a banned word, Midjourney will not allow the prompt. Midjourney users should replace prohibited words with similar but permitted words, avoid using words that are closely or remotely related to prohibited words, or consider using a synonym or other wording.

Forbidden Words in Midjourney

Midjourney has implemented a filter that automatically filters and bans exact or similar words on the banned word list. The list of banned words includes words that are directly or indirectly related to violence, harassment, gore, adult content, drugs, or incitement to hatred. Additionally, it does not allow prompts that include or relate to aggression and abuse.

The list of banned words is not necessarily exhaustive, and there may be many other terms that are not yet on the list. Midjourney is constantly updating the list of banned words. This list is under constant review and is not public. However, there is a community-run list that users can access and contribute if they wish. [0][1].

If a word is on the banned word list or if it is closely or remotely related to a banned word, Midjourney will not allow the prompt. Midjourney users should replace banned words with similar but allowed words, avoid using a word that is even loosely related to a banned word, or consider using a synonym or alternate wording. Midjourney users should always check the #rules channel before submitting their message as the team is constantly updating the list of banned words [2].

Midjourney has a code of conduct that users must follow. The Code of Conduct is not only about following PG-13 content, but also about being kind and respecting others and staff. Violation of the rules may result in suspension or banishment from the service. Midjourney is an open Discord community, and following the code of conduct is essential. Even if users use the service in '/private' mode, they must respect the code of conduct.

In conclusion, Midjourney operates a strict content moderation policy and prohibits any form of violence or harassment, any adult or gore content, as well as any visually offensive or disturbing content. Midjourney has implemented a filter that automatically filters and bans exact or similar words on the banned word list, which includes words directly or indirectly related to violence, harassment, gore, adult content, drugs or incitement to hatred. Midjourney users should abide by the code of conduct and check the #rules channel before submitting their message, as the team is constantly updating the list of banned words.

Updated list of forbidden words

Midjourney periodically adjusts the list of banned words and the list is under constant review. The banned words list is not public, but there is a community-run list that users can access and contribute to. Midjourney strives to provide a PG-13 experience across its entire Service, which is why words and content relating to violence, gore, harassment, drugs, adult content and topics generally offensive are prohibited. The list of banned words is divided into several categories covering the spectrum of topics mentioned above. It is important to note that the list of banned words on Midjourney is not necessarily exhaustive, and that there may be many other terms which are not yet on the list.

Ban and suspension of Midjourney

Midjourney has a strict code of conduct that users must follow. Violation of the rules may result in suspension or banishment from the service. However, it is unclear whether users can appeal a ban or suspension from Midjourney. The sources do not explicitly mention an appeal process or how to contact the Midjourney team about a ban or suspension. It is essential to respect the code of conduct to avoid being banned or suspended from the service. If users have any concerns or questions regarding the service, they can contact the Midjourney team via their Discord server [1][2].

Can Midjourney generate images in specific sizes or resolutions?

Midjourney has specific default image sizes and resolutions that users can generate. The default image size for Midjourney is 512x512 pixels, which can be increased to 1024x1024 pixels or 1664x1664 pixels using the /imagine command on Discord. There's also a beta option called "Beta Upscale Redo", which can increase the size of images up to 2028x2028 pixels, but may blur some details.

Users can only scale to maximum resolution after doing at least basic scaling of an image [1]. The maximum file size Midjourney can generate is 3 megapixels, which means users can create images with any aspect ratio, but the final image size cannot exceed 3 pixels. Midjourney's resolution is sufficient for basic photo prints, but if users want to print something larger, they may need to use an external AI converter to get good results.

How does Midjourney compare to other AI image generators such as DALL-E and Stable Diffusion?

According to the sources, Midjourney is an AI image generator that produces artistic and dreamlike images from text prompts. It is compared to other generators such as DALL-E and Stable Diffusion. Midjourney reportedly offers a more limited range of styles than the other two, but its images are still darker and more artsy. Midjourney doesn't seem to match DALL-E and Stable Diffusion when it comes to photorealism [1][2].

Stable Diffusion is compared to Midjourney and DALL-E, and is said to be somewhere in between in terms of ease of use and quality of output. Stable Diffusion offers more options than DALL-E, such as a scale to determine how well the generator tracks guidewords, and options regarding output format and size. However, Stable Diffusion's workflow does not match that of DALL-E, which groups images and offers collection folders. Stable Diffusion and DALL-E are said to have the same shortcomings when it comes to photorealism, both failing to come close to Midjourney's Discord web app [0].

According to a comparative test by Fabian Stelzer, Midjourney is always darker than DALL-E and Stable Diffusion. While DALL-E and Stable Diffusion generate more realistic images, Midjourney's offerings have an artistic, dreamlike quality. Midjourney is compared to a Moog analog synthesizer, with pleasing artifacts, while DALL-E is compared to a digital workstation synth with a wider range.

Stable Diffusion is compared to a complex modular synthesizer that can produce almost any sound, but is harder to trigger. In terms of image resolution, Midjourney can generate images at 1792x1024 resolution, while DALL-E is slightly more limited at 1024x1024. However, Stelzer notes that the answer to which is the best generator is entirely subjective and comes down to personal preference.

DALL-E is known to produce more photorealistic images, even images that are indistinguishable from photos. It is said to have better understanding or awareness than other AI generators. However, Midjourney is not designed to produce photorealistic images, but rather to produce dreamlike and artistic images. Therefore, the choice between the two generators ultimately depends on the needs and preferences of the user.

How does Midjourney's limited range of styles affect its usability compared to DALL-E and stable streaming?

According to sources, Midjourney's limited range of styles may affect its usability compared to DALL-E and Stable Diffusion. Midjourney's images are considered more aesthetically pleasing, but its range of styles is more limited than that of DALL-E and Stable Diffusion. Midjourney's style is described as dreamlike and artistic, while DALL-E is known for producing more photorealistic images that are indistinguishable from photos.

Stable Diffusion falls somewhere in between in terms of ease of use and quality of results. Stable Diffusion offers more options than DALL-E, such as a scale to determine how well the generator follows the suggested words, as well as options regarding the format and size of the results. Midjourney is compared to an analog Moog synthesizer, with pleasing artifacts, while DALL-E is compared to a digital workstation synthesizer with a wider range. Stable Diffusion is compared to a complex modular synthesizer that can produce almost any sound, but is harder to trigger [1][2].

DALL-E is said to be more flexible than Midjourney, able to offer a wider variety of visual styles. DALL-E is also better at creating realistic, "normal" photographs that would look great in a magazine or on a corporate website. DALL-E also offers powerful tools that Midjourney does not have, such as paint overlay, cropping, and various image uploading, which are essential for more inventive uses of AI art.

DALL-E's model has fewer opinions, which makes it more receptive to style suggestions, especially if that style is less immediately beautiful. Therefore, DALL-E is more likely to provide an accurate reaction to a specific request, such as pixel art. DALL-E also offers a real web application, allowing users to work directly with DALL-E, which can be less confusing than installing Discord.

Compared to Midjourney, Stable Diffusion is supposed to be completely free, making it more accessible to those who can't afford an AI image generator. However, Stable Diffusion is only available as a Discord bot, and users must apply to access it. Stable Diffusion is also considered harder to launch than Midjourney, which is easier to use thanks to its choice of aspect ratio and public gallery. Midjourney also offers AutoArchive, which backs up all images, and a 2x2 grid of saved thumbnails, making it easy to manage the work. Midjourney's Discord app also works better on mobile than DALL-E's website, making it easier to generate images on the go. Midjourney's unique style makes it ideal for quickly generating a large number of pleasing images, without needing to refine the message.

In conclusion, each AI image generator has its own advantages and disadvantages, and each person may have different preferences and needs. Midjourney's limited range of styles may affect its usability compared to DALL-E and Stable Diffusion, but its unique style makes it ideal for generating dreamlike, artistic imagery. DALL-E is more flexible and adept at creating photorealistic images, while Stable Diffusion is completely free and offers more options than DALL-E. Ultimately, the choice between generators depends on the needs and preferences of the user.

Are there significant differences in the quality of the results obtained by the three AI image generators?

The sources do not mention any significant differences in output quality between the three AI image generators (Midjourney, DALL-E and Stable Diffusion). However, the sources mention that each generator has its own strengths and weaknesses, and each may be better suited to different types of images or styles. For example, Midjourney is said to produce dreamlike and artistic images, while DALL-E is known to produce more photorealistic images that are indistinguishable from photos. Stable Diffusion falls between the two in terms of ease of use and quality of results. Ultimately, the choice between generators depends on the needs and preferences of the user.

Tips for choosing the best generator for a specific project or application

According to the sources, choosing the best AI image generator for a specific project or application depends on the needs and preferences of the user. The user must take into account factors such as the type of images he wants to create, the level of detail and realism he needs, the ease of use of the generator, the availability of functions such as painting, the cropping and uploading of various images, as well as the cost of the generator.

If the user wants to create dreamlike and artistic images, Midjourney is the best option. If the user wants to create photorealistic images, DALL-E is a better option. Stable Diffusion falls between the two in terms of ease of use and quality of results. Stable Diffusion offers more options than DALL-E, such as a scale to determine how well the generator follows the guidewords, as well as options regarding the format and size of the results. However, Stable Diffusion's workflow is not comparable to that of DALL-E, which groups images and offers collection folders.

The user should also consider whether the generator is free or paid, and whether it is available as a web app or a Discord bot. Stable Diffusion is completely free and available as a Discord bot, while Midjourney and DALL-E are paid and available as web apps or Discord bots.

Ultimately, the choice between generators depends on the needs and preferences of the user. The user should research and compare the features and output quality of each generator before choosing the one that best suits their needs.

Mid-course alternatives.

As mentioned earlier, Midjourney is a popular AI image generator that creates images from text descriptions. However, it only offers 25 minutes of free render time, which is about 30 images. If you are looking for a free alternative to Midjourney, there are several options you can try.

Here are some free alternatives to Midjourney:

crayon : This is a free and open source solution that offers a good alternative to Midjourney.
GIVE HER : This is another image generator similar to Midjourney and available for free. It is made by OpenAI.
Jasper: This is a free and open source image generator that can be used as an alternative to Midjourney.
Wonder : This is a free and open source image generator that can be used as an alternative to Midjourney.
Invoke AI : This is a beautifully designed image generator with an intuitive interface that can be used as an alternative to Midjourney.
Disco Diffusion: This is a cloud-based text to image conversion system that is easy to use and can be used as an alternative to Midjourney.

If you're looking for something more specific or customizable, Stable Streaming (SD) might be a good option. [3]. However, SD takes more effort to get good results and is not as easy to use as Midjourney. Additionally, there are several other free text-to-image conversion systems, such as Wombo's Dream, Hotpot's AI Art Maker, SnowPixel, CogView, StarryAI, ArtBreeder, and ArtFlow.

In conclusion, if you are looking for a free alternative to Midjourney, there are several options available, such as Craiyon, DALL-E, Jasper, Wonder, Invoke AI, Disco Diffusion, and Stable Diffusion. These systems offer different degrees of customization and ease of use, so you should try several and see which works best for you.

This article was written in collaboration with the team Deep AI et orgs.

[Total: 0 Mean: 0]

Midjourney: Everything you need to know about the AI artist

Midjourney: What is it? Use, Limitations and Alternatives

The process used by Midjourney to generate images from text prompts