Video Project (small one)

[gorillamilk]

Here's 3 frames of 200 that a friend of mine will be using in one of his shorts. It cuts from live action, cuts to CG and pans up to the apartment number, then back to live action. If it looks grainy, it's supposed to match the camera. Granted it's a simple shot technically, but getting it all to match up close "enough" just based on the footage was not. It's my first real attempt at a practical application of mental ray and max other than making wood floors and binder clips. :) Feedback welcome. If anyone has good crit, the sooner the better, so I can implement changes if something is off... it'll take over 30 hours to render it out this weekend. I included a piece of the footage for reference.

[Geek]

Hey!

Everyting looks good so far.30 hours:eek::)?omg how many frames are in a sequence?Maybe it could use a ilttle bit of optimization:)

[gorillamilk]

I had it wrong before. Each frame, at the proper resolution and quality will take about 15-20 minutes per frame. There's 200 frames, so it looks like roughly 66 hours now.

[ransomInK]

wow, that looks great. luv the grain and the lighting for the shot; really blends together.

my only crit would be the bump detail from the wood on the left and right side of the door (door frame). the texture for the bump looks too big...or the bump is too deep. im not sure if it should pop out that much

(if the last pic is the reference shot) i cant see any bump popping out from the frame of the door

[gorillamilk]

You might be right on the grain depth. There is a depth in the original shot that can't be seen in the footage and perhaps I've overdone it a bit. From yesterday I'm 80 frames into the animation, so I think we'll see how it all comes out and make a decision. Here's another shot of the door framing but from a different hallway, lit a different way.

[kadamin]

http://www.youtube.com/watch?gl=PL&hl=pl&v=jofW0jlqPwM

[Ethan_Janssens]

have you thought of doing it in passes?

AO pass seperately is very fast
ZDepth pass seperately is very fast
Diffuse pass seperately is very fast
...

Combined they're VERY long

I'm trying to minimize the nr of passes and I'm at this stage:

Diffuse (without ANY ao in materials)

afterwards use a AO pass (variable ranges) as a multiply
afterwards use a ZDepth pass (variable ranges) as a final touch

for the ZDepth pass i use scanline, no materials, all is grey with the right cutouts, no lighting, no exposure. Just the models actually with the render-elements enabled "ZDepth" in which i can adjust the settings and render it out. A ZDepth render like that (seperate) renders in no time

+ when the diffuse is rendered out (faster without AO nor ZDepth) you can start editing in your editing program and afterwards implement the AO & ZDepth on the already mounted and edited Diffuse pass. You can track the clips back to their frames even if needed.

anyway, hope it sparks something,

regards,
Ethan

[gorillamilk]

Thanks! After some random reading I found, and some previous testing when creating the clip, I realized that what is really killing my render times was all the faces from all my displacement. The scene finished yeasterday and we're taking a look at how/if it fools new viewers and then take proper action from there. The detailed meshes for the displacements might be able to be overcome with a modifier or two from max. I'm going to see if subdivide helps any. If all the geometry was simple boxes, renders would be less than 5 minutes a piece instead of 17-20, which would be nice.

[ransomInK]

that shot with the door framing looks good, maybe its the angle? anyway, would luv to see how the animation pans out- good luck!

[plurbKar73]

Well after doing some of the side Projects, I started a small Video Clip.
My first time doing something like this.
Automobile Promotion like thingy. Will post it here when its done.

Shaggy