Hackster will be offline on Friday, March 1 from 9pm to 10pm PST to perform some scheduled maintenance.

Zach Archer's ePaper Triptych Showcases "Uncomfortable" AI-Generated Procedural Artworks

Using generative adversarial networks (GANs), this self-updating wall piece has never seen a human artist's touch.

Artist and maker Zach Archer has built a triptych with a difference: the three-part artwork is built using color ePaper display panels, and display constantly changing artwork generated by a machine learning model.

"This was inspired by the 'shadow box' art style, especially when differently-sized frames are combined to form a larger piece," Archer explains of the project. "The three ePaper panels have different aspect ratios and color capabilities. The side panels are slightly recessed, which makes their presence feel a bit uncomfortable, as it's difficult to focus on the entire piece (this was an intentional choice)."

Two of the three frames are Good Display 5.79" panels offering red, black, and the background white as color choices; the third, located in the center of the piece, is a larger 7.3" Waveshare display which includes a fourth color, yellow. All three are ePaper displays, using electrophoretic technology to offer a sunlight-viewable image that requires power only when being updated.

Driving the displays is an Unexpected Maker TinyPICO, a compact microcontroller board based around the Espressif ESP32-PICO-D4 module. The artwork is stored on a 32GB microSD card, but none of it was created by hand β€” but rather by generative adversarial networks (GANs), themselves fed using a procedurally-generated prompt list.

"I wrote a prompt-generation script, which generated endless portraits with randomized characteristics," Archer explains. "The best, most emotional images were produced by Stable-Diffusion v1.4, probably because the prompt keywords were tailored for this version. Only about one in 20 generated images was usable, the others were rejected for having multiple heads, misaligned eyes, I just didn't like them, etc. Most of the side panel images were generated on Dreamstudio.ai."

Once every twelve hours the artwork awakens, powered by a 350mAh battery good for 16 days of use, and loads a new trio of images from its microSD card. All the hardware, meanwhile, is hosted within a fake-wood 3D-printed frame β€” appearing from the front as three separate frames but, in reality, a single fused unit.

"This project started in September 2022, when stable-diffusion seemed fun and innocent," Archer notes. "I'm concerned that 'A.I.'-generated images will soon devalue the craft of living, working human artists. My greatest concern is that the images are nothing more than simple style transfers, onto compositions and scenes which are fundamentally boring and lifeless. Soon these uninspired images will be everywhere, and accepted as normal. This is a future I don't want to see, and I'm thinking about this problem every day."

More details are available on Archer's Hackaday.io page; the source code and board design files had not been released at the time of writing.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles