Valve's Source engine to power upcoming animated film
arstechnica.com
arstechnica.com
I do think that there is a ton to be gained from tightly coupled previs pipelines that don't incur rendering overhead. Ideally the artists should be able to see what they are working on in a form that is as close to the final render as possible, but without the wait. Systems that can offer this with low overhead cut down on the amount of time the artists need to get the results they want tremendously, especially in situations where your team isn't big enough to have modeling, texturing, animation, fx and lighting departments that are completely siloed.
In it, he discusses the enormous investment they are making in real-time technologies for future animation projects.
Animation is absurdly labour intensive and highly iterative and, as he says, new real-time technologies will revolutionize the industry.
However, yeah, artist time is more costly, and recently GI pathtracing raytracers have made huge inroads in the VFX industry due to the fact that lookdev/lighting configuration is much quicker for the artists due to being physically-based, so you get very realistic lighting without faking it with AO or virtual lights.
It definitely speeds the iterative process up when the artists can see how the scenes looks on their workstation without kicking it off to the renderfarm and waiting 4 hours for a render to come back.
With Arnold or VRay, it's possible to get a very fast 3/4-bounce GI solution with heavy geometry and textures in a matter of minutes.
In short, Source Filmmaker was a never officially released tool that was used as part of the production of the TF2 and L4D promotional videos. It provided tools for recording footage, basic editing, and advanced rendering that went beyond the engine's realtime capabilities.
I haven't used Source Filmmaker but I wouldn't recommend using the MovieMaker tool for actual films. The use case is completely different.
I'm curious if you actually need a license to use an engine in a film. Presuming you produce all your own assets and otherwise have an appropriate license for the engine to render it once, is there any precedent for requiring a license to perform it as well? I'm assuming you don't need a performance license for NLE software like premiere or vegas?
And with that thought I felt my third eye open just a crack, revealing a vision of how movie/game franchises of the future might be funded and marketed. :) Interesting stuff, or at least, in a few years it might be.
That said the difference between the rendering of the artists workstation and the renderfarm is getting smaller and smaller. At some point folks like Weta will be able to put together pretty kick ass effects for the dailies. Its pretty good now but eventually ...
I can see how Source would be a good way to make a reasonable-looking animated movie on the cheap. But with a reasonable amount of money to spend, then either:
a) You're reusing enough existing stuff (character models, physics enginey stuff, et cetera) that you can't possibly spend that much money, or
b) You're re-doing enough stuff from scratch that there doesn't seem to be much advantage to doing it in Source rather than something more conventional like Maya [or whatever they're using nowadays]
Who knows what the break down is - and whether the choice to use source is based on cost efficiencies alone. Perhaps they are going for a particular visual style... I would have thought the source engine is sufficiently old enough to give such a film an interesting retro aesthetic...
So yes, a Pixar movie ($200m) is an order of magnitude more expensive than this film ($18.6m).
It can take a week just do 3 seconds of animation and all the labour is, of course, exceptionally talented and extremely expensive.
Not doubting you, but do you have a source for this?
Puss in Boots, for example was $130 million and Kung Fu Panda 2 was around $150 million. So you're looking at 1.25m to 1.5m per minute.