A project by Phil Van Treuren
The much-loved impressionist painter Claude Monet was born in 1840, and lived long before the development of computers and artificial intelligence. His artwork is particularly well-suited, however, to use in a new generation of AI animation tools — thanks to the unfocused, dreamlike qualities of his most famous paintings, their organic, natural settings and his use of vibrant colors and spatial imprecision.
Generative AI, Dreaming & Impressionism
When I started experimenting with early generative AI image creation programs like DALL·E from OpenAI, I quickly noticed that the visuals they produced had a lot in common with dreams.
In the first generations of these programs, the images generated from text prompts usually had a dreamlike quality of “almost right, but not quite.” Hands would have too many fingers; written text would be nonsensical; subjects would sometimes warp into strange mutations that still had an oddly logical, unconstrained “flow” to them from a dreamer’s perspective.
Because of the “canvas of dreams” quality that these early generative AI images had, it was a great tool for producing impressionist artwork that you might see from Monet, Renoir, Degas and others. There’s nothing that feels “unnatural” about an impressionist oil painting that has imperfect spacial ratios, blurry horizons and messy brushstrokes.
Newer, more advanced versions of image generators like Stable Diffusion are now able to produce photo-quality pictures that are almost indistinguishable from real life (if that’s what you want them to do). They’ll also create static visual art in any style that you can discribe in a text prompt. Gone are most of the dreamlike morphs and mutations that were common in early AI images.
Video Generation: A New Playground for Artificial Intelligence
AI video generation is the new kid on the block now, and it’s interesting to observe that these first attempts at text-to-video has the same “dreamscape” feel that early text-to-image had. In fact, the short moving clips that we’re starting to to be able to produce with generative artificial intelligence have an even more dreamlike quality to them.
As before, this means that these new (and still sort of primitive) AI video tools are fun platforms to use for experimenting with content that has an impressionist feel. (Trying to get them to animate realistic, photo-quality stuff produces occasionally impressive stuff, but it the results are more often nightmarish.)
That’s not to say that this technology won’t get much better . . . and very quickly, too. Right now you can only generate three-second video clips, but I’d be surprised if movie-length stuff isn’t possible in a few years.
To get these results, I used a new tool from PikaLabs that’s still in closed beta testing, but the technology is already really impressive — and really fun to mess around with. You can simply give it a text prompt and see what the AI generates from text alone, or upload your own image for it to animate (as I did with these Monet paintings).
I chose animations of 43 different Monet paintings to include in this video. It was a fun project, but I wouldn’t say it was quick or easy to do. Some of them took dozens of attempts to finally get a three-second animation that looked good enough. (You’ll notice that in the clip of a train station, the locomotive is actually pulling into the station backwards.)
The piano music I used in the video is Erik Satie’s first Gymnopedie.
I know there’s a lot of animosity toward generative AI from some artists, writers and actors, and I can certainly understand their point of view. I wrote more about my thoughts on the topic in this article, which I hope you’ll check out if you get a chance.
Regardless of how you feel about artificial intelligence, I hope you’ll learn to embrace it and explore how it can help you live a more creative and fulfilling life. It’s not going away and will only become a bigger part of the human experience as time passes, after all.
I’ll add one more bit of info, which I hope you’ll take to heart: almost all of the research, editing, video creation and writing for this project was done on my cell phone, with my thumbs, in my spare time.
No one taught me how to do any of it, but I had to learn to use some technology that I wasn’t comfortable with a year ago.
You can do this kind of stuff too, regardless of your age or experiences. Embrace uncomfortable change, use it to become better, and don’t let anything about the future intimidate you.