Over the past few weeks, the Bright Side of Life has been exploring the possibilities of Artificial Intelligence. In our earlier blog, we researched ChatGPT. This latest blog describes some of the possibilities offered by Midjourney AI, a web service that generates images based on natural language descriptions (“prompts”). In our opinion, the results are impressive.
A San Francisco-based independent research lab exploring new mediums of thought and expanding the imaginative powers of humans has developed Midjourney. The team is led by David Holz (Leap Motion, NASA) and advised by several respected industry players (Apple, AMD, Tesla, Intel, GitHub, Avid, Second Life) who have already made their mark in the industry.
You access Midjourney via a Discord Server. Discord is an American VoIP and instant messaging social platform. Users communicate with each other or with robots via voice calls, video calls, and text messaging and exchange media files in private chats or as part of communities called “Discord Servers”. After creating a Midjourney/Discord account and logging in to the Discord Server, you can use Midjourney to create new images using natural language text commands (“prompts”).
All you have to do is “ask” Midjourney
You ask Midjourney AI to produce a design by writing a description of what you want using English language commands. Midjourney can create almost anything you can think of, so you can design a new logo or create a new design for a website with just a few carefully chosen words. If all this sounds too good to be true, it is! Unfortunately, Midjourney delivers results as flat images and can’t as yet create a vector image (Adobe Illustrator), a layered image (Adobe Photoshop) or a functional website. So, once Midjourney produces your design, you will still need to implement the design using the website development and visual design tools with which you are familiar.
Every design starts with a visual idea. You can base the design idea on something you have seen or imagined. You can even add an image and use it as a starting point for the design. To set Midjourney’s task of converting your idea into an image, you must first describe your vision. Inevitably, the quality of your description will determine the quality and accuracy of the design. However intelligent, the Midjourney AI server is merely a computer, so you get what you ask for.
This is an example of working with Midjourney:
After entering your description, Midjourney will create four images, each representing a slightly different interpretation of your commands. For example, the following prompt produces four designs in a single image. You can later request a larger version of one or more of these designs.
“Create modern material web design homepage for IT company, using images of server rooms, computers, people, in colors white teal orange”
Designs such as this can be a good starting point for discussions with a client. You can quickly generate page designs and other elements so the client can give feedback. You can use this feedback to modify your description, creating new designs for a new iteration with the client. Once the client is satisfied, you can use the final designs as the basis for the website development.
Building a website using Midjourney designs
We frequently use two Visual Builder tools to develop advanced WordPress websites: Divi and Elementor. Both provide drag-and-drop editing interfaces and a wide range of modules that we can use to create and manage website content. We believe the developers will soon integrate their tools with AI to help build working websites automatically from AI-produced designs.
In this tutorial, they use two AI tools — Midjourney and ChatGPT — to help construct a homepage design in Divi. Firstly, they use the Midjourney bot to create the visual design. Then they use ChatGPT to write the copy they’ll use. Finally, they recreate the design, using the visuals from Midjourney and the copy from ChatGPT in Divi.
Creative blog banners using Midjourney AI
For our earlier blog about ChatGPT, we created the banner using Midjourney. Since we use wide images for our banners, we also need to instruct Midjourney to adjust the image’s aspect ratio to 16:9. The prompt for this image was:
create modern material design about artificial intelligence, worldwide, business analysis, chatgpt, showing floating code blocks, ai, lines, arrows, earth, face in colors white blue gold, UHD –ar 16:9
Midjourney first provides a grid image containing four interpretations of your prompt. Several standard buttons are also displayed directly below the image.
If unsatisfied with the results, you can re-run the command with a single click (circle with arrows). In our case, the new set of images looked like this:
Furthermore, four buttons exist to create a new version based on one of the generated images (V1, V2, V3, V4). By clicking one of these buttons, Midjourney will use the corresponding image as a source for four new variations.
When you finally see an image that matches your design idea, you can request a higher-quality version by clicking one of the four upscale buttons (U1, U2, U3, U4).
One drawback to these generated images is that everyone can now easily create an image to support almost any story – even if the story is an invention. Of course, such an action can be harmless, but some may be tempted to take advantage of these features.
For example, there is the question of who will play the new James Bond. And since the announcement that Daniel Craig would quit, there have been rumours that Idris Elba will take the role. You can generate such a story yourself with ChatGPT and generate a suitable image to accompany your fake story using Midjourney.
Let’s try the following prompt:
Idris Elba as james bond holding a gun
There is no need to provide a reference photo when requesting a design that includes well-known people. Midjourney has enough online reference material to create something. By adjusting the format, resolution and quality of these images using commands understood by Midjourney, it’s straightforward to create fake photos for posters or to accompany a (fake) press release.
Initial conclusions about Midjourney AI
Midjourney AI is a tool that can create almost anything you can imagine. It can help you when you are in discussion with a customer about the design and the elements the customer wants to use. However, using Midjourney AI to generate details like “the page title should be 10 pixels bigger and 15 pixels more to the right” is challenging. But Midjourney AI will definitely be one of the tools we plan to use for development and web content production.
So keep watching our blogs. A later blog will go into more detail concerning many more possibilities offered by Midjourney. And if you’d like advice or want to apply the latest techniques within your company, website or other services? Then feel free to contact us.
Finally, a little tip. You can use your own face to generate designs so long as the image you wish to use is publicly available online. Once you have a suitable photo online, you can make yourself into a hero with just a simple command:
make me a superhero