According to the developers, the gallery was created using a combination of VQGAN, a neural network capable of generating images that look similar to others, and CLIP, OpenAI's neural network that is able to determine how well a caption matches an image, as well as Unreal Engine 5. The engine was used to render 360° images at various locations, the images were then fed to the networks with unique prompts, and after that, the AI-generated paintovers were re-projected to their original locations.