A few weeks ago I got invited to beta test the new Adobe Firefly. I have had some pretty interesting results. The first thing that I learned is that the AI doesn’t seem operate in a conversational style like ChatGPT. For instance, if you ask it to do very specific things with images it has created, it can’t seem to do that without redrawing with completely new results. Unfortunately, the majority of the time those results are completely different. I know that it’s still the beta, but the simple concept of just undoing or going back to a previous iteration doesn’t exist (or it’s really well hidden).
That said, some of the art Firefly produces is quite stunning and beautiful, even if by accident. Here are a few of my favorites, along with the prompt that got it there.
Hero(ish)
Prompt: please draw a super hero in a clown suit, sitting on a cloud eating cotton candy in Tim Burton style.
Styles: graphic,steampunk,chaotic
Video Game Cut Scene
Prompt: Please draw a city under siege by an army of blimps
Styles: steampunk, cyberpunk
Eye‘m A Tyrant
Prompt: please draw a floating eyeball with arms as a super villain, in the style comic books
Styles: Concept art, Marble, Chaotic, Antique Photo, Graffiti
And my personal favorite so far was my attempt at creating something that felt like Basquiat.
Bustiat
Prompt: draw a robot, with a boombox for a head in the style of Neo expressionism.
Style: chaotic
Sure, if you look at the prompt, the art is clearly not a robot with a boombox for a head. I honestly believe where this technology is today is more about a compromise or partnership between the thoughts in your head and the translation of those thoughts by the AI. Sometimes it’s what you asked for and sometimes it’s not. Either way, if you’re open to it, the results can be a beautiful piece of accidental art.
As the technology gets better, I’m sure we’ll be able to fine-tune the art as much as we’d like. I personally am not sure that’s a good thing.