Since we wrote our first post less than a month ago, our team had been experimenting with the VQGAN + CLIP Google Colabs that were freely available and the newly released DALL-E 2. In just a few weeks, things have dramatically shifted.

Now that almost anyone can access Midjourney, the depth of AI art experimentation has grown quite rapidly. It’s easy to get sucked down a prompt-design rabbit hole and I found myself doing this for 5 hours straight yesterday. Below are images I was working on to re-imagine Disney dark rides in concept art form. Midjourney is must have a pretty good model of Disney and it’s rides, as these came out pretty wild.

Something I realized while working with all of the AI generative tools is that the models use many existing artists work, old and new. You can type in almost any artist name and they’ll try to match the look you’re going for. This is one of the concerns I’m sure many designers and artists will have. At some point, when generative art like this gets good enough to look real, how can artists protect their rights and profit from copycat AI? The images below all share the same prompt “Time traveler droid by Syd Mead”. The results are stunning. I have always been fascinated by Syd Mead’s work, so this was really fun. But I could never call this my own.

I can see fear in the eye of many types of artists and designers to what this may become. What happens when AI can generate full 3D environments, fine tuned and optimized 3D models, product packaging design, fashion, jewelry, etc. My hope is that artists will use these tools as a starting point to enhance and explore new ideas they may not have thought of before. Although if I wanted to write a comic book, creating the artwork is kind of doable already. The models are already great at comic book style superhero characters in almost any style. All of these took just minutes to create.

Much of my time has been spent coaxing the AI to design me a time machine. Not just a visual representation, but with actual build plans, control panels and circuitry. If it can get even close to something that looks good as a concept for a movie prop, good enough for me. Plus, it’s fun to see what it can come up with.

Fashion! Who knew that an AI could create jewelry and handbags. Also, the ‘photography’ is quite amazing. Both Midjourney and Dall-E 2 were put to work here. These took more time to drag into the image gallery than to create them with simple prompts.

I am very much looking forward to seeing what happens in the coming months. As more and more of us mere humans use these tools and the AI knows what we like, it will only get more realistic and we won’t be able to tell what is real anymore. We all hope to keep our day jobs as the team that builds the physical things, solders the electronics together and programs the software, but AI will probably take that over soon too.

My vision of the evil AI coming for our jobs

Author: Brian Dressel

AI Art Experiments - Part 2

Doritos Behind the Scenes

Water Simulations in SideFX Houdini