Midjourney – a neural network that generates pictures from a text description
Nowadays, few people are surprised by a neural network, these things can process video, conduct a dialogue with a person, search for materials on the Internet, write music, recognize objects in a photo, help process photos, and much more. Today I want to talk about the grid of the drawing picture – midjourney.
The Midjourney team positions itself as an independent research laboratory dedicated to expanding the creative abilities of mankind.
The project was launched in February 2022 by scientist and entrepreneur David Holz.
David Holtz is a graduate of the Department of Applied Mathematics at the University of North Carolina at Chapel Hill. As a student, he worked at the Max Planck Institute, where he studied neuroimaging algorithms, also worked at the NASA research center and took part in the development of LiDAR technology (detection and ranging using light, the technology is used, among other things, in iPhone cameras).
More information about the founder’s personality can be read this post here.
Midjourney is an independent, self-funded project with a team of 11 full-time employees and a large number of external consultants. The list of all project participants is available on the official website of the project.
The work of Midjourney is provided by speech recognition algorithms and algorithms that create images. Simply put, Migjorni can recognize printed text and convert it into pictures. To do this, you need to describe the plot in English, send it to the processing grid and wait for the result. After the result, you can upgrade a little, increase its quality and download.
You get pictures like this.
The grid is presented in the format of a Discord bot. The invented plot should be sent to the bot in the chat, and in response it will send a generated picture.
The presence of such a tool in the future will greatly simplify the work of the designer, solving the problem of the lack of content on stocks, the long search for the necessary materials.
To quickly create content, you need imagination, the ability to structure information and interpret it into understandable phrases. Well, that’s cool!
I am sure that no neural networks and Skynets will replace the designer, but they will greatly simplify his work.
Now about how to test this miracle of technology. 😊
Now the grid is at the stage of open beta testing, anyone can get access. For this you need:
1. Go to the Migiorni website – midjourney.com and click the “join the beta” button.
2. The site will redirect you to Discord. There you need to log in or create an account. Authorization through Dis is required – the neural network works in the format of a Discord bot.
3. Next, you need to give the grid the necessary permissions.
4. After that, the system can direct you to Midjorni’s personal account. It looks like on the screen. To get to the Discord bot from your account, you need to click on your avatar at the bottom of the screen and select the “Go to Discord” item. It can immediately send to the bot, it happens in different ways. 😊
5. The opened channel of the neural network will look like this.
6. You can find help information in the “welcome” section, it briefly describes what’s what.
7. After that, you need to go to any channel whose name begins with newbies: “newbies-102”, “newbies-132”, etc.
8. Being in the channel, you need to enter the command “/imagine promt” in the text line and the text of the plot, it should turn out like on the screenshot.
9. After the text is inserted, you need to press enter and you can watch the process of generating pictures in real time. This will take about a minute.
10. You can not wait – when the picture is ready, the neural network will send a notification. You can go to the entry in the channel through the notification center of Discord.
11. After reviewing the result, you can ask the grid to work out options for one of the plots by pressing one of the buttons V1, V2, V3 or V4. The buttons correspond to one of the four proposed options.
12. Buttons like U1 etc. necessary to initiate the process of improving the quality of the picture – “upscale”.
13. After the picture is finalized, you can download it or modify it a little more, increase the size to the maximum, and also evaluate it.
There is a more extensive networking guide on GitHub. Various additional commands and settings are available in it.
Link to guide here.
I should also add that most products are made by people to make money and Miggiorni is no exception. To test the operation of the grid, developers provide “25 actions” for free: you can generate 25 sets of 4 pictures in low quality or fewer sets and for the remaining points (actions) work out options or increase the quality of images.
When the limit is reached, the bot will send you this notification.
To increase the limits, you can go to the network admin panel and purchase one of the offered packages.
That’s all for me. I highly recommend testing the grid and really looking forward to when neural networks become one of the design tools, as familiar as Figma or Photoshop. 😊