LogIn
Technical Artist and AI enthusiast Sean Simon has shared the results of yet another experiment with Stable Diffusion, showcasing one more potential use case for Stability AI's open-source application. This time, the artist demonstrated DiffusionCraft AI, an unusual combination of Minecraft with SD, which allows one to turn Minecraft builds into visually appealing concepts.
According to the creator, a Python script feeds a Minecraft spectator cam live capture through Stable Diffusion, the image-to-image features of which then turn a bunch of blocks into an image in real-time based on the given prompt.
"While most img2img videos are done in post, this is quite possibly the first to run it live on Minecraft, comments Sean. "This is intuitive because it creates an inspiration feedback loop where I get inspired by what Stable Diffusion interprets, and I continue to shape it down a certain path. It currently runs on a seconds-per-frame ratio, but this speed will only get better as the hardware improve."
You can learn more about DiffusionCraft AI here. Also, don't forget to join our Reddit page and our Telegram channel, follow us on Instagram and Twitter, where we share breakdowns, the latest news, awesome artworks, and more.
This pack can help you get that realism that 3D often lacks with various different finger smudges captured from real surfaces.