AI-generated art has been big news lately, as several purpose-built deep learning models hit the internet over the last couple of years. While they can produce some genuinely cool artwork when properly guided, they often make images that are so bad they become entertaining. But one can get a new piece of art by typing in a few words and waiting a few minutes, which means the options are practically limitless. To display the good (or funny) AI artwork, Alex Thiele built this little desktop E Ink easel called Portrayt.
E Ink screens are still much slower than just about every other display technology on the market, but they have serious benefits. They’re very readable in bright light and only consume power when refreshing, making them extremely efficient. Some newer models can even display colors instead of just monochrome or grayscale. The model that Thiele used is a Pimoroni’s Inky Impression 5.7” 600x448 seven-color E Ink screen. It takes 30 seconds to refresh, but the color range is great for showing artwork. 7 colors isn’t a lot, but software dithering can mimic other colors.
This E Ink screen is a HAT for Raspberry Pi single-board computers. All Thiele had to do was attach the screen to his Pi and plug in a power supply. Those sit on a little wooden easel, which people can find at their local craft stores. Most of Thiele’s work went into the software, which generates AI artwork based on a prompt and then pushes the image to the display.
The software generates art through the Stable Diffusion API (Application Programming Interface). Users can enter prompts manually or let the system automatically update the display at timed intervals. When generating a new piece of art, the user can enter a single prompt or combine two different prompts and select one of the animation frames. OpenCV then downscales the image into seven colors and the software shows the converted image on the E Ink screen.