using reduced 5d cone engineering with transdimensionality between 5D and 4D allows for a global curved space 2d global picture body with localised control and to adapt such data along in one direction of time.
so you reduce the curved model to something which both looks good per frame when the model is upscaled in curve and that scales optimally with the dimension of time.
here you can reverse data dimension flattening of old video you know those more blocky artifacts sometimes more subtly in the picture.
you can model the data to archetypes natural archetypes like humans landscape archetypes. an the less important accuracy of noise is in the picture with such detail looking more like for less exact data can be reduced.
Put this together with advancements dynamic lossy data compression. Remember these models would be far more a perfect fit too space time would have infinite resolution and would allow for better specific feature upscaling of old 3d graphics as well.