In this article, we'll show you how to create your first piece of art using Midjourney's artificial intelligence and tell you how to get the results you want.
Before you start using the Midjourney neural network, you need to take a few preparatory steps.
Step 1. Join the Midjourney Discord server
If you go to midjourney.com, you will notice that there is no "picture generator" on the site. This is because the bot is hosted on a Discord server. If you don't have a Discord account, create one.
You can click on "Join Beta" or "Sign In" on the Midjourney homepage, and you'll be prompted to sign into your Discord account.
Step 2. Read the rules and accept the terms and conditions
Once you're on the Midjourney server in Discord, you can go to either the "Getting Started" or "Rules" channel to familiarize yourself with the community guidelines and learn how the Midjourney neural network works. Be sure to accept the terms of service or you won't be allowed to use the bot.
All new members get a trial run: you'll be able to make about 25 requests. Why is the number approximate? Because the trial is based on GPU minutes, not real images. The fact is that some queries, scaling and variations require more GPU time than others.
The trial period will allow you to generate images with AI and learn the basics before upgrading to a paid plan. Generating images can be tricky, so you will need time to practice.
Step 3. Add the Midjourney bot to your server (optional)
You can add the Midjourney bot to your Discord server. To do so, click on the bot's profile image on any channel where it is active (for example, one of the Newcommer Rooms channels). You will see a brief description of the commands and an "Add to Server" button. Click it.
A new pop-up window will ask you to grant Midjourney Bot access to your server. From the menu, select the one you want to add the bot to, then click "Continue".
Navigate to the server you selected. You can now use commands and queries to start generating images with Midjourney. Type "/imagine" to initiate the invite command, and add your query.
Step 4. Enter the queries to generate the image and wait for Midjourney
To generate a picture, use the "/imagine" command.
In the editable text field, enter a request for the bot. To help the neural network better understand what you want to get in the picture, write queries in English. For this example, we wrote "Australian shepherd dog with heterochromia, photorealistic, portrait style". Give the bot as much detail as possible. Midjourney will then generate four variations based on your query.
Below the results, you have the option to request:
U- zoom, which will provide a higher resolution version of the same image;
V- will provide variants of the same image.
Step 5. Download the image generated by Midjourney
If you are happy with the resolution and variants, you can download the generated image.
There are two main ways to download pictures from Midjourney.
1) Open the original picture in Discord
You can download the picture directly from your Discord feed. Click on the variant you want to download, then click on "Open Original" in the bottom left corner.
This will open the image in a tab on your web browser, where you'll have the option to save it as a lossless, high-resolution .webp file.
If you want to download the image in PNG format instead, use another method.
2) Download the image from your Midjourney profile
Log in to your profile on the Midjourney website and go to the Home page.
On this page of the site, you will see all the images you have created and can view them individually. When you click on an image, you will see the prompt, related tags, resolution information, parent image, and sharing options. Midjourney, like other neural networks for drawing, is evolving and gaining popularity. And in the near future, we will be getting the most satisfying generation results from artificial intelligence to meet our requirements.
Click on the three-dot menu below the image to access the "Save Image" option.
How to write queries to Midjourney's neural network for best results
Midjourney has a helpful article on "Tips for text prompts" that we recommend reading. Here are a few of them:
1) Define image parameters
Midjourney's neural network defaults to a 1:1 (square) aspect ratio for all images. If you don't want to crop or resize the image, indicate this to the bot in language it understands. To do this, use the "--ar" tag in your request.
The Midjourney neural network defaults to a 1:1 (square) aspect ratio for all images. If you don't want to crop or resize an image, specify this to the bot in a language it understands. To do this, use the "--ar" tag in your request.
You can also specify a specific width and height in pixels using the "--w" and "--h" tags, but using "--ar" is currently better supported.
2) Explore other users' pictures on Midjourney's community showcase
The community showcase on Midjourney is a great place for inspiration. You can hover over any image to see what queries the user used to generate the picture.
We recommend finding images that you like with the elements you wish to reproduce in your images. Then read through the queries and see how different users came up with those results.
3) Specify specific artists and/or art styles
If you wish to generate an image in a specific style, think about what words can be used to describe it, such as "modern," "detailed," "surreal," "photorealistic," etc.
If you want the art to mimic a particular artist, add "by artist name" to the query.
In the screenshot above, we wanted an impressionist piece in the style of Claude Monet, so the query says "woman walking through a rose garden at sunset, Claude Monet, romantic".
4) Try to use common words and phrases
Midjourney can create completely different images. The neural network associates certain words and phrases with the images it was shown during training. Therefore, it will be more effective if you include words in your query that can be clearly identified visually . The more images displayed in Google search, the more chances Midjourney will correctly interpret your clue.
Take for example the clue "Australian Shepherd Dog with heterochromia, photorealistic, portrait style". If you look at the results, none of these images show a dog with heterochromia, a condition in which an animal or person has different colored eyes.
This is probably because the word "heterochromia" is visually undefined. There are not enough images on the internet for Midjourney's neural network to replicate this.
If you encounter a similar problem, try describing the meaning of the word. For example, we changed our query to "Australian shepherd dog with multicolored eyes, photorealistic, portrait style."
Finally, among the retrieved images there is a variant with one blue and one brown eye, which is a classic manifestation of heterochromia.
News From
CrocoappsCategory: Mobile App Developers Profile: Crocoapps has been developing games and apps for over ten years. We especially love projects on React Native, Swift (iPhone), Unity 3D / C#, Unreal Engine 4 / C++, WebGL, Android Studio. We provide a turnkey application development service of the full cycle - from assistance in the development of technical specifications to promotion in the network. Leave a request by email: This email address is being protected from spambots. You need JavaScript enabled to view it.
