Nice. For mixing photos one could use existing frameworks such as Libvisual or maybe hooks used by VJ software such as LiVES (there is a lot more VJ software for *NIX). The point here is that you're basically making a video stream out of static photos using automation from metadata (EXIF, XMP, user defined) and effects to 'mix' them together. The process is similar to VJing, just the target audience and produced work is not. For example, this demo is mostly a video (made using LiVES ~6 years ago). Can you imagine the output being less video/motion_picture (due to input; more photos) but still with nice effects to blend from one photo to another?