Hello,
I've completed the first shot of the openning sequence :



Color grading may evolve a bit once final editing will be completed.
I've encountered two issues for this shot : one regarding camera projection precision, the other one regarding the fog limits.
I've splitted the scene in two area, one for the near Greeble City Blocks, the second for the far Dystopia City Blocks. Both city blocks are from Daz3d that I've imported into iClone through 3DXchange. I've created ambient occlusion maps for each block, for wich I had to create a second set of UVs through Softimage|XSI.
I've processed 20 Greeble blocks and 111 Dystopia blocks. Not all blocks are used in this shot, but at least now I've them all available if needed.
The scene area is pretty big, it extends to ~70000 units from the center of the world, the second area starting at ~40000 units.
The first issue I had was a camera projection precision error. This was expected, actualy. When you get this kind of error, the geometry start flicking randomly. The simple workaround for this is to fine tune the camera range. This setting is not only usefull to 'see through' objects or limit the far visibility, but also for this kind of precision error. So I've setted the near range to ~2000 and the far range to ~7000 ( you have to divide by 10 for this setting, not sure why, this sounds to me as a consistency issue regarding units ). This way, no more flicking, problem solved.
The second issue I had was the odd limit to 20000 units for the Fog far limit.. Not sure why such an arbitrary limit, but I've discovered it while setting up the depth pass. I was screwed because the last blocks were at ~70000 units.. To solve this kind of problem, you usualy have to scale the entier scene, except in iClone, you can't scale the camera.. So I had to create a dummy object, tranfer the animation of the camera to this dummy object, parent this dummy object under the parent of all the city blocks. Then I removed all camera animation ( transform only ) and link the camera to the dummy object. After that, I could scale the global dummy object by 10 and I was set with a nice Fog gradient from black ( near ) to white ( far ).
I've also discovered that the Fog works per surface ! Not even per pixel or per vertex, but per surface, which means if you have a big flat box with other objects on top, the flat box will have a constant fog level while the other objects will have their own fog level. In the case of the depth path, this is an issue because you can have a surface nearer than some other objects more white than those objects, wich is wrong. I've managed to split some objects but not all. This is a surprising issue, nowadays any realtime engine use at least per vertex fog if not per pixel.
As expected I've spent a lot of time preparing the assets, nearly 2 weeks, and did the shot itself in two days ( including setup, animation, rendering and compositing ). Now that I've those assets ready, the next shots of the openning sequence will be faster to create.
Cheers,
Guy.
--
guy rabiller | GOETZIWOOD STUDIOS
"N.O.E." (Nations Of Earth) Sci-Fi TV Show, Showrunner.