NEURAL-GENERATED Botanical Prints (2022)
Among the first pieces of art my wife and I hung on our walls were several reproductions of 18th century botanical prints – beautifully painted images of flowers, fruits, and ferns, drawn to scale with incredible attention to detail, and which bordered on photorealistic.
Inspired by these prints and by the work of Rob and Nick Carter on transforming still lifes, I teamed up with British designer Matt Kevan to create a very special kind of botanical “print”. The species represented in the print have never existed, but are instead generated by a neural network. And the print itself transforms from species to species at an imperceptibly slow pace throughout the day.
The print, of course, isn’t a print on paper. The botanical images are displayed on a white-balanced and luminance-matched display, similar to the one I created for A Canvas Made of Pixels years ago, wrapped in a beautiful custom-made antique frame with a linen mat.
Here’s the morphing botanical print, sped up 20x so it’s possible to see how one species shape shifts into the next:
The basic approach to creating the imagery – and, I want to emphasize, Matt did all the work here – was to train StyleGAN2 on a few thousand botanical prints and then use the model to generate a number of synthetic species, smoothly interpolating between them. After upscaling the imagery, I used AfterEffects to slow down the video to the point where motion was below the threshold of perception and to add paper texture, shadows, and some patina.
The effect is delightful. Every time I walk by the artwork, it’s different – sometimes subtly so, as when a plant has grown some new berries or slightly longer leaves, and other times, quite dramatically so, as when a species has transformed into another entirely.
Here’s a 30-minute video at 1x (read: super slow-mo) speed and complete with ~30 different synthetic species, for anyone who wants to roll their own digital synthetic botanical frame, or have something neat in the background while working.
December, 2022