Dof

[glorybound]

Any suggestions for DOF in post? Also, I don't thing I fully understand the ZDepth pass. Do you set the min to 0 and the Max to the furthest distance from your camera? Or, do you start and stop with min/max to where you want the DOF to be?

[Joelaff]

I like Frischluft. It generally works well. There may be others for stills, though.

As far as the channel goes. If you use a 32bit floating point (EXR) depth channel then it is not as critical. However, it is still a good idea to set the min and max to reasonable values. I usually temporarily turn on the clipping planes in the camera to get a visual idea of what to set these to. Keep in mind you need to encompass your whole animation if you animate. YOu don't have to have the min at zero either. It can be right in front of your closest object.

Note that a 16 bit int format has significantly more resolution in terms of discrete steps then a 16bit float. So a 16bit int format is better than a 16bit EXR, though a 32bit EXR is best. But... the question comes down to, what does your comp app support? Many of the blur filters (including Frischluft I believe) actually only use 8 bits worth of the values for blurring, but you can use other tools in floating point or 16bit int to get the depth info prepared before feeding it to the blur plguin. You want the black to white range as tight as possible around your objects in your scene.

I would save both a filtered and an unfiltered depth pass as well. Sometimes one works better than the other. Also, note rendering double res and scaling down helps a lot with quality (or even just rendering the depth pass at double res, scaling the RGB up, blurring and scaling it back down).

[super gnu]

you can leave min at 0 and set the max to the furthest point in your scene.

that gives whatever plugin a nice greyscale range for the whole scene that it can use. you then adjust where you want the blur in the post plugin.


ive not done dof im post for ages.. when i last did it frischluft lenscare was the best around.. somebody mush have done better by now, that was years ago.

[joconnell]

You set the start at the closest thing scene in your camera view - there's no point in getting depth values for bits of your scene close to camera that are out of shot. You set the max to the furthest away thing that you see in your camera view.

The idea is you're making yourself a pass that captures an even range of close to far that suits the depth of your scene - there's little point having a really far away max value if it goes way beyond what's actually rendered in your scene, likewise there's no point in cutting it short so that parts of your scene all end up getting the same level of blurriness.

I take it you're working with this in photoshop which makes correct min and max values more important. If you work with floating point depth channels you don't have to set min and max, it'll automatically capture the range of your scene (unless it's fucking huge) but again you'd need a zdepth plugin that can work with float values. All other zdepth plugins use a ramp from black to white and you have to be careful to choose min and max values that contain your scene as that's what vray will use to set where's black and where's white.

With the depth pass, you have a look at the scene and figure out what shade of grey you want to be the sharp point. The depth blur will then make everything both closer and further away from this value get more and more blurry.