How to use Midjourney: a simple guide
Miscellaneous / / April 05, 2023
We tell you how to create pictures for free, "cross" them or stylize your own photos.
The Internet is replete with various collections of works by Midjourney, a neural network that creates images by description or even brings together several pictures in one. If you also want to experiment with artificial intelligence, we will show you how to do it.
How to access Midjourney
You can use Midjourney through a bot in the Discord messenger. The bot is available to all users, but with a limit on the number of attempts. So far, there are only 25 of them per account, so do not use generations for nothing - both the creation of new pictures and the improvement of the received works are taken into account.
Here's how to access the neural network:
- Download Discord and launch the application or open the web version of the service in a browser.
- Log in or create an account if you've never used it before Discord - You will need an email address and phone number.
- After authorization, go to the server with Midjourney using a special invitation link.
After clicking on the link, you will be taken to a public chat, where dozens of Midjourney's works appear every minute on user requests. Your pictures will also be displayed here. You can also view them in a separate window. To do this, click on the "Mail" button in the upper right corner of the interface and select "Mentions".
Midjourney can also be added to your server, where there will be no strangers. To do this, you need to click on bot in the list of users on the right and select "Add to server".
The place where AI is used does not affect its operation in any way, so you can limit yourself to a public server.
How to set up Midjourney
There are Newcomer Rooms channels on Discord with newbies chats. That's where the pictures are made. Select a chat with any number, click on it and write a command /settings. Next is input.
In the set of options that appears, select the ones that are highlighted in the screenshot above:
- Engine version. We click on the last one, MJ 4 - the newer, the better. MJ Test is two experimental images instead of four, MJ Test Photo is an engine with a bias towards photorealism.
- Image quality. To wait less, click Base quality (base).
- Styling Level. We put the average, Style med, so that neural network not "reimagined". Results with high and very high may be something too far from the query.
- Image Enlargement Algorithm. Choose Regular upscale - standard. Light upscale increases the resolution of the image, but adds less detail, and Beta allows you to get twice the size of Regular, but it is unstable.
- Access mode. We put Public mode, public, because private, Private mode, - paid.
- Processing mode. Here, too, only Fast mode is available - instant with a generation counter, and Relax mode without it is paid.
- Remix mode. It will allow you to combine pictures and work with the finished image - you will need it a little later.
Next, type the command /imagine, press enter, write any word into meaning promt and confirm the submission.
A message will appear with the terms of use of the service - click on the green Accept ToS button.
This completes all the preparations. Now you can experiment through the same command /imagine, which is automatically completed with the value promt. Here it is enough to type the necessary values for generation.
How to generate image in Midjourney by description
This is the simplest scenario, when you enter the words that create a picture. Use English. Generation is carried out in real time and takes a couple of minutes - four options will be created right in the chat. For example, here is the result for the query /imagine prompt life hacker red fox.
The four options received are not yet final. Below them will be the U and V buttons. U allows you to increase the resolution of the selected picture. V - create similar options for one of the four pictures. The arrow button is a new generation (remember the limit of 25 attempts).
After selecting an image for increase (for example, by pressing U1), it will appear in the chat. Below the image are the following buttons:
- Make Variations - four more new images, similar to the one you enlarged and examined in detail.
- Light/Beta Upscale Redo - variations in size increase (with less detail or with higher resolution).
- Web - opens a link to the image on the Midjourney gallery site, where all your generations will be displayed.
Generation parameters (words) can be separated by commas. The more of them, the more difficult it will be for the neural network to get into your request, but sometimes even 5-6 values can be taken into account and combined. Below are examples as requested ultra realistic, neon, demon, mortal kombat, japan style And end of the world realistic 3D.
The Midjourney bot is quite omnivorous, so you can experiment with different characteristics of items and objects, supplementing your queries with more precise generation parameters.
How to create an avatar in Midjourney
To work with the finished image, you need to call the settings through /settings and turn on Remix mode. This will allow you to use Midjourney to style your photos on avatar. To do this, after /imagine prompt, just specify a link to the photo. For example, on the page of the social network "VKontakte" or on any open hosting, and then specify the necessary parameters.
For example, after the link, you can specify star wars style (in the style ofStar Wars") or marvel character (Marvel character). Here is an example from Star Wars with an old photo from VKontakte.
Comparison with the original.
You can also set styling using view characteristics. Sonic from movie, Like Gandalf, Simpsons art style and so on. The last two examples are in the images below.
How to "cross" images in Midjourney
Remix mode is also required here. Further, it suffices to indicate after /imagined prompt two links to images. It can be both heroes of films and video games, and, for example, two vehicles - that's enough fantasy.
1 / 0
"Crossing" the Predator and the protagonist of the film "V for Vendetta"
2 / 0
Gimli + Mario
3 / 0
Hybrid Bugatti Chiron Super Sport and SMZ S-3D
How to correctly formulate a request for Midjourney
Correctly composed request guarantees that Midjourney will try to accommodate your requirements. If some words or introductory words are not understood by the neural network, then it will ignore them. To make the query as accurate as possible, in addition to simple characteristics, you can use parameters.
Options
There are a number of parameters that can be set directly when creating a request with /imagine. They are written through two hyphens no space right after promt and words to generate an image. With their help, you can change the settings on the go.
Here are some of them:
- --w — sets the width of the image in pixels;
- --h — sets the height of the image in pixels;
-
--ar
— sets the aspect ratio of the image, for example --ar 16:9; - -no - indicates what should not be in the picture, for example “no fire” - no fire;
-
--styleize
- changes styling with available values: 625, 1250, 2500, 20000 and 60000; -
--quality
— changes the image quality with values: 0.25, 0.5, 1, 2 or 5; - --chaos <0 to 100> - changes the randomness of the results, the higher the chaos, the less similar the pictures will be to each other;
- --stop - stops image generation;
- --test --creative - adds creativity to the results;
- --wallpaper - allows you to get a high-quality image with high resolution;
- --video - allows you to get a video of the generation process - you will need to respond to the generated emoji image with an envelope.
Usually characteristics and parameters are separated from the request by two colons with spaces on both sides " :: ». This is a kind of query syntax that is considered correct and allows you to visually separate characteristics from parameters.
Example: /imagine promt russia future:: --chaos 100. Or: /imagine promt darth vader:: like pink panter --test --creative. The results are below.
A set of parameters can also work without any distinctions in the request - even if everything is specified through comma.
Adviсe
The process of building any query can be divided into three stages:
- First you need to decide on the main object that you want to style. It can be a person, an animal, a thing, a plant, a food, a country, whatever. Also, there can be several objects.
- Next, the object or objects are complemented with the details that you want to see: colors, textures, environment, style, external analogy, and so on.
- Then the generation parameters are set, which allow you to specify the size of the images, their randomness, quality, and others.
Example: /imagine promt santa claus (1) with dirty work pants and repairing a car (2):: --quality 0.5 --chaos 20 (3). The result is below.
For help in making requests, you can use special service, which offers a large set of "working" words and settings.
Also, if you enter the typing string "/", you can open a scrollable list of commands. Among them are:
- /info - shows basic information, including the number of generations used (Lifetime Usage).
- /help — help with the basic capabilities of the bot (on English language).
- /show - when specifying the work ID, it shows it in the chat if, for example, you lost a picture in the user generation stream. ID can be found at Midjourney website when logging in via Discord: open the image, click on the three dots, then select Copy and Job ID.
How to save the generated image
If you simply click on a picture in the chat, it will open with an "Open original" link so that you can conveniently save it in a large format.
You can also request that pictures be sent to your Discord private messages. To do this, you need to reply to a post with the result from the emoji bot in the form of a letter.
Well, do not forget about the previously mentioned Midjourney website. If you log in there through your Discord profile, you will get access to the gallery of all your generations and their variants in full resolution.
How to bypass the 25 request limit
Midjourney has a paid subscription for $10 per month for 200 generations and an unlimited subscription for $30. But pay them with Russian bank cards you can’t - you need a foreign one.
Therefore, if necessary, you can try to create a second Discord account and get another 25 attempts, but this will require a second phone number, since it is tied to the profile.
Read also💻🏞📱
- The neural network showed the characters of "Harry Potter" based on book descriptions
- People will have digital twins. This is how futurists see the future
- The neural network showed a video with the history of the Earth - from the birth of the universe to the future of mankind
- TEST: Neural network or artist? Determine the author of the picture
- High-quality photos of Van Gogh and Napoleon: neural networks restored the appearance of historical figures from their portraits