Yeah probably they should have gone with another "sharper on next gen" cross gen effort and delayed this a year to give it the polish it needs.
Resolution bitching aside they ran into a huge problem here. Instead of making characters look better they went the route of increasing their number hoping to please people with that idea. NeoGaf is unthankful as ever although they always coped with worse graphics on GTA games (when did characters get discernable fingers?) because of open world.
The excuse seems not valid here. But then, Ubi is not innocent teasing us with prerendered movies that don't show the problems of a realtime rendered world. And the problem is not resolution or missing shadows (many of the screens show the shadow quality is pretty good, they are nice and soft without suffering of low shadow map size).
The problem is low light situation where no visible shadows are cast. While the environment has (25GB?) of prerendered AO (which actually shows) the player misses some basic contact shadow to make him not float. Characters in the shadows look like characters in the sun, just a little darker. They would need much more AO to reflect the ambient lighting situation. Without that they look washed out, which pronounced by their low poly models.
Having mass scenes with hundrets of characters on screen is impressive, but it is a bad idea to use the same low poly models when there are only a few (dozen) people on screen. Yeah, it makes the game look more homogenous, but maybe prerendered mass cut scenes would be a better decision than having those scenes make the game as a whole look worse.
And it probably runs slower because of these. Increasing character count by factor 10 for the mass scenes sure might increase AI and even simple collision detection calculations exponentially bringing the CPU to its knees. But in 90% of the game situations the CPU is probably idling around with a handful of AI calculations.
_/¯ , what ya gonna do?
