G-buffer out channel??????

[asasasddd]

G-buffer out channel??????

how do you have V-ray it kick out a Z-depth image?

[egz]

Make sure the Z-buffer is selected in the G-buffer rollout.

[586]

Also, you can save the render in RLA or RPF format for use in a compositing app like Combustion. I've used a standard Z-buffer output through the Max Effects rollout, but it doesn't seem to antialias very well.

[Da_elf]

here has been some problems i found with faking dof. if i have a mirror with frosted design on it and i want my focus to be on the mirror. then the G-buffer stops at the mirror. but the things reflected in the miror should be out of focus but since the g-buffer stops there then you cant do that. so try a fog and that will allow you to do it. however even though this works, with glass windows it wont since a: g-buffer doesnt go through glass. and b: max has problems with fog and opacity

[586]

That's a really good point you have there. The DOF effect doesn't take into account the apparent distance to objects reflected in a mirror. So, here's my question:

Take a scene with foreground object and a mirror placed away from it. Set focus on the object so that the mirror itself is out of focus. Does this also mean that the reflected image of the foreground object is out of focus? Or, do you get a blurry mirror with sharp reflections?

[Da_elf]

naturally
the reflected image is actually twice the apparent distance away
on a piece of paper draw two dot to represent the plan view of where your eyes are. then a circle for your ball. then a line fo the mirror plane, then another circle where the ball appears to be in the reflection. for the ball to be in focus you need to draw lines from the eye dots to the ball. where they intersect is in focus. now then. keep drawing those lines. they cross eachother then continue on to the mirror. they are diverging. and when they reflect, they diverge even more. so the apparent image is not in focus. to have that in focus you draw the lines to the image. however in reality, the lines are not in focus at the mirror. but they are converging. when they reflect they are still converging. and finally when they meat. they are at the back end of your ball.

[586]

OK, that's what I thought. SO, adjusting the DOF in post should affect both real and reflected objects in such a scene. Or, at least that's how it should work in theory, no?

[Da_elf]

well. that was my idea of using fog for the z-depth since it reflectes in the mirror. using black materials with white fog to get the depth. the g-buffer will trace to the mirror. recognise it as an object. then stop there. it wont trace the apparent depth in the mirror.
next point is glass. the same as a mirror the g-buffer traces to the glass and stops there. but with the fog. there is a glitch in max with fog ive found and opacity. Which is why i can never use opacity mapped trees or leaves when using fog in a scene. So for reflective shots fog works. but for glass your screwed either way.
vray however works fine in either case as it should

[asasasddd]

nothing happens.....

when I select the Z-buffer is selected in the G-buffer roll out nothing happens.....
even when I save the render...
any sugestions...

[Da_elf]

are you saving as an rla or other format that used the z-buffer? and when you say save do you mean using the save output. or just clicking the little disk icon in your VFB?

[zerofractal]

I found something quite interesting regarding the zbuffer in Vray:

I made this render as a box. the main idea was to map it onto a box and to produce a VR, but I wantred the VR to have DOF on it (wierd heh!)..

I clipped the range form 3000 (white) to -3000 (black)


The image is a zbuffer fom a box render in Vray. The interesting part is that the front view gradients towards black, and whats on the bak of the camera (labeled as back) gradients towards white. Looks very cool, the problem is that for any post production DOF effect the zbuffer must be consistent form the camera to the target, but Vray goes ahead and does negative units to the fornt and positive to the back.

To my understanding the cure for this was very easy. positive z values negative, and thats it a 300 should be -300 and a 50 should be -50. So I worite a simple script that takes a rla and inverts the values of positive zdepths per pizel so they are all negative.

To my surprise, the ending Zbuffer image turned out as:



Looked goos, but therr was a very wierd gradient in the side views where things went from positive to negative in the original image, making the post effect worthless.

Analyzing the original image, i found the following:



Somewhere on the sides I has a zDepth equal or very near to 0, and it was physically impossible that the distance was 0 form the camera to the objects whose pixels correpsonded the gbuffer corresponded to.

SO if Vlado gets any time to tweak the GBuffers to solve the reflection he can probably tweak this. There are two chances either give teh correct depths in the sides, but maybe pass from -400, for example, to 400 in the neighbour pixel, or just keep the distance to teh camera regardless of the direction.

I suppose this applies to any camera type with a wide angle over 135 degrees or so...

[trixian]

Hmm...if this issue is fixed, one could render the image out in unclamped colours (32-bit float maybe), and add the z-buffer and output it to a .EXR image. Then all one needs is someone to write a plugin to light ones scene whilst taking the z-buffer channel into account.

*rubs hands gleefully*