In September we already reported on slashCAM from an AI model, which can create a 3D Ken Burns animation from a single image (!!).
Various technologies are combined and partly optimized. Among other things the Z-depth estimation is combined with an object separation to get defined edges and not to let objects "tear" in the depth. The AI also helps to conclusively supplement missing image contents that are exposed behind objects in motion.
So far there was only one paper by Simon Niklaus describing the model. Since the research was done with Adobe, some permissions had to be obtained before he could release the source code. The latter happened four hours ago on Github.
So if you want to create your own 3D Ken Burns effects from single images (and know how to set up Python environments) you can find the whole PyTorch project for 3D Ken Burns Paper on GitHub. Nice thing...