I am working on a large interior scene for a new gymnasium animation for USC Long Beach. The scene will have moving cameras and moving people (rpc’s and full 3d models using mocap files)

The gym has 280 high output downlights. I know that this will me way too time consuming to render 7200 frames so I was wondering what other light setups I could use to get good results for animation purposes.

Also does anyone have any input on holding down render time for large scenes in animation.

How do movie studios do it?

Would breaking my scene into xref files make the scene any less time consuming at render time?

SOME SPECS: Max 2009, MR engine, Intel Quad core processor, 896,000 polygons

Thanks

studios probably use network rendering via backburner with 3ds max

Bare with me on this but I believe this will handle your problem.
First nail down what rendering time is viable - even if it is within a certain scope - let’s say - 10 minutes a frame. Then work out what must be in your piece to make it come off as an artistic piece, as it is not just the lighting but of course it’s inter relation with materials and effects in your scene. Then when you have your assets created at the correct LOD and are now lighting and tweaking materials - you have maybe 6-7 options depending on your needed quality level you have established from your first step above.
The different lighting options can be effects by: 1) GI solution. Sometimes FG is faster then photos and vice versa and sometimes faster together. If you want to go super cheezy you can go to standard lights and use large attenuation distance omni lights to “fake” the far reaching of FG. Hell you can even put an omni at 0,0,0 in your scene and in the ambient slot of the light put a Ambient Occlusion Shader and then just add accent lights. It’s really first doing what I found in a great article on art once:
“ART is a word which summarizes THE QUALITY OF COMMUNICATION”
“It therefore follows the laws of communication.”
“Seeking perfection is a wrong target in art. One should primarily seek communication with it and then perfect it as far as reasonable. One attempts communication within the framework of applicable skill. If perfection greater then that which can be attained for communication is sought, one will not communicate.”
“Example: a camera that shoots perfectly but is not mobile enough to get pictures. One must settle for the highest level of technical perfection obtainable below the ability to obtain the picture.”
Then working back from that product - then you can ask your specific questions because at that point you would have named your product. Even look at the reports from the lead staff who did Iron man - they new that they needed to nail the burshed steel and that that was going to be the bug on render times and what would really make their shot come off, so what they did was handle that problem to make their shots come off within their scope of work and then viola.
Did that help...?

I would start by generating your overall ambience to the scene. If the gym has windows, then you should use the daylight system to bring light in. If you do not have light coming from the exterior, then start with a few lights only, instead of trying to render all 280 lights. You may have to fake it a bit. You can still make it look like all 280 lights are turned on by using self illuminated materials on some of the bulbs, and photometric lights on others. Also consider have lights with shadows turned off. This may help you a bit for render times. Experiment with using FG bounces and see how much this can help light the scene.

I agree with the comment about network rendering. If you have multiple machines then you should use distributed bucket rendering or backburner to split the images into strips.

I find that breaking the scene into xrefs is helpful at render time, depending on the complexity of the model.

Good luck

I don’t think xrefs will help at render time. Their purpose is to help scene management, but at render time, the renderer will receive all the geometry no matter what.
mental ray will optimize the load of geometry, but if you want to help it, use mr proxies for the repetitive objects (benches, light fixtures, etc).

I’d suggest to avoid 280 lights. Group them and use less if possible. Don’t use area lights, use point lights. If you need to use area lights, make sure the shadow sampling is low (like 4) not the default of 32.

Lights (shadows) and materials (reflections, blurry reflections) are the most time consuming processes. Focus on those.

If your windows use sky portals, lower the sampling to 8.

You can save the FG map if there are no animated objects (camera excluded). You can create and update the FG map every 5-10 frames before rendering and will save you tons of rendering time when you do the full sequence.
Remember, Photons + FG (0 bounces) is usually the fastest solution, and if you can use a saved FG map because you have no or little animated stuff, remember that you don’t need to save the photons map, as the FG map will include their effect.

Compositing is another way to save time if you think the rendering might have to be redone if things need to be changed or fixed. Just plan it well.

But it all depends on the quality you consider acceptable. If more optimizations make the images look bad, then the only solution is to throw more hardware at it.

I like the suggestion from Rodrigo about Art. Maybe a good looking even if not realistic image is something that can be considered, but be aware that it can be time consuming to create a good Non Photorealistic look when using photorealistic rendering engines.

Rob

I use Mr proxies to get render times down and it does have a noticeable effect, I make animations of large warehouses, which are produced using a single machine with twin quad core processors, it would be impossible for me to produce these animations at the size and length I want, using near photo realistic lighting. So its a compromise, I spend time on the materials, I use to give me the effect / look I want without pushing up the render time, I stick to relatively simple MR lighting where ever possible, limit the number of lights I use and which lights produce shadows etc. As this is an animation, the fact that some lights don’t cast shadows, or are not even in the scene can easily never be noticed by the viewer, as I try to make what’s happening in the animation hold the viewers attention, and not the lighting or lack of it or cheats hold his attention.