Timur Ozdoev talked about turning Midjourney-generated images into assets for the Cognition Method game and shared his thoughts on the relationship between AI and traditional art.
My name is Timur, I worked on Cyberpunk 2077 as an Environment Artist at CD Projekt RED, and I am currently working on an exciting GTFO in the 10 Chambers family. In our free time, my friends and I are creating the game Cognition Method, where we experiment and put the most important things for us into it.
We have a small team and we were looking for any smart tools that would help us make the job easier. Text-to-image tools are very popular, and we decided to try them in action.
I know that our producer director Slava Lukyanenka played with DALL-E, but I didn't try other options: when I got Midjourney for the first time, I generated pictures non-stop until 5 a.m. It was a lot of fun.
I think that each AI generator is good in its own way and is in no way inferior to the other.
Midjourney is very easy to use and gives very fast results, it was perfect for us to quickly test some ideas and find inspiration. You can use it on your phone while you are on the subway.
In general, the time Midjourney takes to generate the result you need largely depends on how you read the generated pictures. Sometimes we see some meaning and inspiration in the most unexpected places. Midjourney opens a portal to such places. Sometimes long, sometimes fast, usually by accident, just keep going! Even a messy and random picture can give you serious ground for the idea that you will grow in the game. Sometimes I can go through different combinations of words for hours, but I can't always completely rely on the generated image alone, for me, it’s a good first step toward creating a concept.
Turning AI-generated Images into Game Assets
Pete Mc Nally did a great job converting an AI image to 3D!
I will briefly show you how we approached this on the example of a game creature in Cognition Method.
At each level, the player will encounter a “shadow”. We conceived it as a disproportional silhouette of a person, which is a projection from another space.
Midjourney concept: “human in the void”:
Having found the right words, we got exactly what we wanted, and I set up several model variations in DesignDoll, where you can very quickly create any human-like form you need.
Then I went to 3ds Max/ZBrush to make them look very similar to the concept:
Our good friend (and the tech lead) Andrew Indrikson prepared an effective shader in Unity, so we just added a texture straight from space and this is what happened in the end:
Text-to-Image Tools in Gamedev
I generally like the idea of talking to a computer in human language, this is definitely the next level of our interaction with programs! I’m sure that such progressive products as Unreal Engine and Blender will definitely pick up this initiative in the future, and others will gradually follow them. Not only text-to-image but also text-to-anything.
AI VS Traditional Art
I'm very optimistic here: it's always great when technology allows you to create complex things in a very simple way so quickly. AI will definitely affect the life and work of specialists, I think that the industry will see a lot of new and interesting changes.
But, of course, it is not enough to just have super tools or be able to draw technically to make interesting art – for artists, in my opinion, it is more important to feel and understand. AI is a great power. But great powers in various forms have always surrounded mankind, and only a few have mastered them.
It will be good if masters of traditional art add new technologies to their arsenal and make this world change again! All these tools open up new horizons for everyone without limits.