An Endless Outpainting Tool Based on Stable Diffusion
Peter Garrity, Leonidas Deligiannidis · 2023
We developed a tool that utilizes the capabilities of machine learning to explore the possibilities of image outpainting. Our tool is built around an image generation model based on Stable Diffusion and provides an easy-to-use command-based interface. The command-line interface allows us to generate an endless set of images. It starts with a given image, and after the outpainting process, it feeds the generated image as input to the next iteration. To preserve the original image's dimensions, before feeding an image as input to the next iteration, we scale it down so that the generated images are the same dimensions as the original image. By using several iterations like this, we can repeatedly create new details around the edge of an input image.