Additional Feature Ideas

It would be great to incorporate the sort of markup featured on Place Atlas. Instead of just marking up the final image, users could mark up the image throughout time.

Performance is poor. The main grid shader can easily be optimized in a lot of little ways. One big way would be for the million base quads on the grid to rotate their tessellation towards the camera, so that they would only ever need to draw one side quad, instead of two. This way, the max vert output of the geometry shader could be reduced from 11 to 7.

Once performance was improved, it would free up bandwidth for some additional render features like depth-of-field (which could be tied to the highlighter tool) and cast shadows. It may be possible to write a very cheap shadow caster by baking out some of the data in the bake phase.

Alternatively, the application would be much more accessible if people didn't have to do the 3 hour bake step. In theory, the bake could be eliminated by generating a subset of keyframes, and then calculating the inbetweens at runtime off of the diffs from the keyframe. Combining the bake and the viewer systems into one would make things get much more complicated though.