Final Words… on After Effects 3D

Now there’s a thread on After Effects CS6‘ raytrace 3D every week, but since we all have to move on eventually and focus on more positive things than getting a heart attack because a big company made a mess, let me summarize some of my thoughts that we discussed on the mailing list, forums and elsewhere.

First, the most obviously glaring flaw with the whole matter is that it builds on the program’s outdated infrustructure and thus is confined and severely limited by its paradigms. I’ve many times written that in order for any of this to have meaning, we need a real 3D world, not a composition and in order to move through this world, we need a better way to navigate and animate our stuff, meaning the whole thing be better speedy, have some nice 3D gizmos and the best graph editor you can imagine. For that, nobody is asking to reinvent the weel. To date, I’m perfectly content with OpenGL 2.0 as long as it offers a reasonably close representation of the final result within its technical bounds and the textures don’t look all too shabby. I do not even need fancies like viewport shadows, though that is possible these days.

The second important point is that Adobe themselves should not get involved in providing the final render engine, but only the API infrastructure for third-party vendors to hook into. This may sound contradictory, given the buzz they made about the raytracing stuff as one of the standout features for CS6, but eventually this will come back to haunt them. Why? The "serious" 3D world has long moved on beyond plastic-y 1990s raytracers such as the one we have now. People use global illumination and physical shaders every day now and as it stands, Adobe pretty much has only 2 options. One, they settle for this smallest common denominator and face the same issue they were facing with the old 3D system again in 10 years – users getting frustrated because the stuff no longer holds up to then current standards or the rigidness of the system limiting again development on other ends. Don’t get me wrong, you can bet your mom’s favorite bra on that we will get support for importing geometry created in a 3D program, displacements, camera mapping and a few other things, but at some point the development team will have to decide how far to go because the other alternative is equally bad. In order to keep up, they would have to invest so many resources, that eventually other things will suffer and users will get equally frustrated. In an odd way, it could be a lose-lose proposition for them, after all. Therefore the reasonable thing to do is to just forget the whole CS6 ever happened and move on to enhancing the underlying stuff, so that tools like AtomKraft can fill the gap.

Finally, all of this better be reasonably performant. Say what you will, CS6 is just completely terrible, even if you have a graphics card that works for the raytracer as a means of acceleration. Contrary to what some people like Steve Oakley claim, it’s not a matter of buying yet another graphics card. If he can, then that’s good for him, but there are enough hard-working people that need to scrounge every penny to get a new computer every few years. That and of course the armies of Mac users with their limited choice of hardware in the first place and the other big group who have perfectly running systems, sometimes just a few weeks old and were shut out simply because Adobe does not support their AMD/ ATI cards or they run on an unsupported NVidia card even. What makes this even worse is really the fact that you can go out there, buy a moderately priced core7i system, throw a 3D program on it and render most of the stuff in pure CPU code just as fast…

%d bloggers like this: