Any given frame is just looking at something from one side though, this is the case for video games as well and it’s part of the reason why real time rendering is so much slower. It’s an art and game direction challenge to make things look good however you want to not a technical limitation (in the sense of, you can make a video game look like a Pixar movie does today, it’s just going to render at days per frame instead of frames per second)
There isn’t really a conceptual difference between rendering a frame with the intent to save it and later play it back, and rendering a frame with the intent to display it as soon as it’s ready and dispose of it.
Toy story 1 took days to render a single frame, now it could be rendered on a single home GPU at 24 fps no problem, which would be real time rendering.
To clarify my first paragraph. The challenge is not that it is impossible to render a video game with movie like graphics it’s that the level of effort is higher because you don’t have the optimizations, and so art direction needs to account for that.
As far as considering unexpected behaviors, that is technically only a concern in psuedo-nondeterministic environments (e.g. dynamic physics rendering) where the complexity and amount of potential outcomes is very high and hard to account for. This is a related issue but not really the same one, and it is effectively solved with more horsepower, the same as rendering.
I think the point you were making is that potentially, artistic choices that are deliberately made can’t always be done in real time, which I could agree with. Something like ‘oh this characters hair looks weird the way it falls, let’s try it again and tweak this or that.’ That is awarded by the benefit of trial and error, and can only be replicated real time by more robust physics systems.
Ultimately the medium is different, and while they are both technically deterministic, something like a game has potential for unwanted side effects. However, psuedo-nondeterminism isn’t a prerequisite for a game. The example that comes to mind are real time rendered cutscenes. They aren’t fundamentally different from a movie in that regard, and most oddities in them are the result of bugs in the rendering engine rather than technical impossibilities. Similar bugs exist in 3d animation software, it’s just that Hollywood movies have the budget and attention to detail to fix them, or the luxury to try again.
I’ll end with, if we have the Pixar server farm sufficient hardware, there is nothing that says they couldn’t render Luca or whatever in real time or even faster than real time.
Any given frame is just looking at something from one side though, this is the case for video games as well and it’s part of the reason why real time rendering is so much slower. It’s an art and game direction challenge to make things look good however you want to not a technical limitation (in the sense of, you can make a video game look like a Pixar movie does today, it’s just going to render at days per frame instead of frames per second)
There isn’t really a conceptual difference between rendering a frame with the intent to save it and later play it back, and rendering a frame with the intent to display it as soon as it’s ready and dispose of it.
Toy story 1 took days to render a single frame, now it could be rendered on a single home GPU at 24 fps no problem, which would be real time rendering.
To clarify my first paragraph. The challenge is not that it is impossible to render a video game with movie like graphics it’s that the level of effort is higher because you don’t have the optimizations, and so art direction needs to account for that.
As far as considering unexpected behaviors, that is technically only a concern in psuedo-nondeterministic environments (e.g. dynamic physics rendering) where the complexity and amount of potential outcomes is very high and hard to account for. This is a related issue but not really the same one, and it is effectively solved with more horsepower, the same as rendering.
I think the point you were making is that potentially, artistic choices that are deliberately made can’t always be done in real time, which I could agree with. Something like ‘oh this characters hair looks weird the way it falls, let’s try it again and tweak this or that.’ That is awarded by the benefit of trial and error, and can only be replicated real time by more robust physics systems.
Ultimately the medium is different, and while they are both technically deterministic, something like a game has potential for unwanted side effects. However, psuedo-nondeterminism isn’t a prerequisite for a game. The example that comes to mind are real time rendered cutscenes. They aren’t fundamentally different from a movie in that regard, and most oddities in them are the result of bugs in the rendering engine rather than technical impossibilities. Similar bugs exist in 3d animation software, it’s just that Hollywood movies have the budget and attention to detail to fix them, or the luxury to try again.
I’ll end with, if we have the Pixar server farm sufficient hardware, there is nothing that says they couldn’t render Luca or whatever in real time or even faster than real time.