Basically his approach is to render all the 3D models of the same shot as single png images and then layer them on top of each other in 2.5D.
Is this the way most CGI movies are made? I’ve always thought 3D Animated movies at Pixar had the entire 3D scene rendered.
Of course in VFX you need the 2.5D approach, but my personal experience is by this method, your run into a problem with lack of cohesive cuts.
Say, for example, you have a CGI environment with mountains and stuff. A full scale environment as if it was a video game.
When you place the camera at difference angles, to get the different shots, its cohesive, it feels like it is indeed the same scene (which it is).
But in 2.5D, each new camera angle needs to be made from scratch. Say you rotate the camera 180 degrees, you need an entire new scene, different lighting etc. When you composite all this, the two shots feels like they are from two different movies, so how do professional CGI/VFX studios get around this problem?
This is not how CGI movies are made, your assumption is correct. You can easily find scene breakdowns from Pixar showing how they use a full scene, moving through light and polishing passes. Why would CGI movies be made this way? This is a “Compositing” tutorial, not a “making a CGI movie” tutorial
In fact, some people like to render only one image and work from there.
Separating in layers can allow more geometry in the same image.
If you do a lot of render layers and render passes , you can set the final look in compositing, and some things are easier to do there.
Also, when you separate layers, let’s say you want to change the sky, it’s easy and only one quick render to do. This can even be done in the compositing software by importing the camera.
But, you can also set the final look in 3D, render everything in one image and use comp only for little tweaks, it’s another way of working.
Especially for animation, where you don’t have to match a filmed plate , or integrate actors that can make a lot of sense.
Nowadays, you can use rendered motion blur and DOF, which some time ago was done in a post process. So I think now it can make sense to try to nail the look as much as possible and render one image.
What can be interesting to have is render passes, rather than renderlayers :
Like cryptomattes to isolate some parts, reflections to tweak them. You can even separate some lights and tweak them in comp.
Which render passes to output ? : really depends on the case, render what you need, don’t feel forced to render something because others do
So the look of the shots are known from the start, lighter try to match it as much as possible, and the compositing tries to get closer. They also copy-paste presets from shot to shots.
In the very end, there is a color grading pass to make sure every shots matches in continuity.
The best is to think about what you need, and how you’d like to work. And also where does it makes sense to finalize the look. If you work like for SpiderVerse, I guess you need a lot of render layers and render passes. The final look is done in comp.
For another project it may be better to do most of the work in 3D. And of course that can change within a project. Some sequence may require heavier compositing.
A good read on the subject :
Basically both approach can make sense , it’s part cultural, and also tied to a production, render times, final look ect…
CG movies are routinely constructed in multiple “passes” that are then composited together. The renders of each “pass” are undoubtedly done at different times. This gives the production team the chance to build the final shot, instead of committing everything to “one grand roll of the dice.”
“Building the shot” requires more planning, and more post-production work, but it also offers flexibility. (And: “if a render goes wrong,” there is less to have gone wrong. The “dice that you roll” are now considerably smaller.) You still have options. The “thing that went wrong” isn’t the whole shot.
