PDA

View Full Version : Lenscare & Vray


Kinematics
07-07-2011, 05:46 AM
I am using Maya2010 with Vray and Lenscare.

I wanted to know about how to work best with all three to get the best DOF possible and reduce artifacts from occurring. Based on information from @cgbeige, I reckon I would just determine the furthest point and object is from the camera and use that as the white value and nearest as the black value. Is this correct? Or does having an expression control the near and far plan a better idea?

Next up, how do I render a depth map at 16bit png and do I need it to be in 16bit for best results or is an 8bit png enough? I want to render the beauty pass as 8bit and the zdepth pass at 16bit while just hitting the batch render once. Is that possible?

Thanks in advance

Hamburger
07-09-2011, 07:15 AM
In render settings, click "Image Format Options" button, a new menu will popup and you can change png to 16bit there.

Ideally you'll probably get best result with a 32bit exr which I think lenscare supports but you can try 16bit if you want.

Kinematics
07-09-2011, 07:54 AM
Hey Hamburger, thanks for the reply. The main issue right now is that when I render in png 8bit, it takes like an hour a frame. At 32bit exr, it jumps to 4-6hours.

So I was hoping to render the beauty as an 8bit and the depth pass as 32bit without actually setting up a new render just for the depth pass. I assume it would waste more time rendering that way.

Some people mentioned rendering the depth pass at 32bit double resolution without anti-aliasing. That gives better results. is this true?

I was looking at Alex Romans stuff and the DOF is flawless. I see problems all the time in my DOF when I apply lenscare. Is the background being full black a bad idea? As in distance infinity or should I always put a non visible sphere to clamp the back off?

Hamburger
07-09-2011, 08:31 AM
Hey Hamburger, thanks for the reply. The main issue right now is that when I render in png 8bit, it takes like an hour a frame. At 32bit exr, it jumps to 4-6hours.

I think that is unusual. I always though V-Ray, internally, did it's calculations in 32bit and just spat out an 8 or 16bit image in the end so there was never a difference (it seems that way when using the V-Ray framebuffer). I could be entirely wrong here though.

Maybe you should contact Chaos because it is quite a big difference.

Some people mentioned rendering the depth pass at 32bit double resolution without anti-aliasing. That gives better results. is this true?

Well, technically the depth buffer shouldn't be aliased because if you were to alias the pixels between edges, it will be "artificially" filtered and produce the wrong depth value that the lens software expects.

If you are to double the resolution without anti-aliasing like you mention, you'd have to do it to the beauty as well and not anti-alias the beauty pass, so both images match each other.

It seems Alex Roman used filtered zdepth pass for his cameras in 3rd/7th though. If you look on his Vimeo with HD, you can see errors/filtering issues in some shots its not perfect but excellent nonetheless. I guess a lot of it disappears when encoded.

Not sure about clamping though, if you were to render in 32bit this part wouldn't be an issue.

CGTalk Moderation
07-09-2011, 08:31 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.