Enlarge / A portrait of Commander Keen from Commander Keen 6 upgraded via AI. (credit: frigis9)
Last night, a Reddit user named frigis9 posted six images that featured detailed graphical upgrades to classic MS-DOS computer games such as Commander Keen 6 and The Secret of Monkey Island. The most interesting part is how they did it: by using an image synthesis technique called “img2img” (image to image), which takes an input image, applies a written text prompt, and generates a similar output image as a result. It’s a feature of the Stable Diffusion image synthesis model released last week.
In response to a question from another Reddit user, frigis9 provided more details about how he crafted the image using a front-end program called Visions of Chaos—a versatile app that provides access to many different styles of generative computer art: “For nice, simple portraits like these, I set the init image strength to 0.25. For darker, less-detailed images, may have to crank it up to 0.35 or 0.4. You’ll just have to fiddle with the image strength value, and once you’ve got that (i.e. checking to see if your output image more-or-less looks like your init image), work on the prompt to add or remove details. Adjust, rinse and repeat about a hundred times till you get the perfect (or meh, good enough) image.”
Over the past week, users of Stable Diffusion’s img2img feature have upgraded children’s drawings, turned scribbles into gleaming knights in the desert, upgraded their profile pictures, and more. At the moment, running img2img on your own machine is still a somewhat technical process, but by using graphical front-ends like Visions of Chaos and models on Hugging Face, you can experiment with it yourself. Development is moving quickly, so better ways to use them are likely just around the corner.
Read 1 remaining paragraphs | Comments