arseniytestin.com

Composite render is the only render you want. 

Really, there is no point in CGI if what you get is a flat picture that costs ten times a photograph and has no extra value. We want a render to retain in 2D, retouch, all of the freedom we had in 3D, production stage, so we go composite.

Think twice, though, for it is not a novice's tool and you have to ask yourself, first, if you are skilled enough to take advantage of it. A poorly managed production render can otherwise be overkill for your deadline and morale. 

Composite render Does imply extra time. Extra time to visualize why and how to split the image. Extra time to physically split the final file in components. Extra strain render time wise. A new set of skills to master a compositing software. If you manage to get there alive, is harvest time. Hopefully you had a few ideas for the retouch and you can finally go wild.

Being smart, instead, saves huge resources before the production render kicks in. Visualize the Final Image. Ooooh! Beautiful, isn't it? Well, you know half of it comes from retouch, so only half the time can now be allocated for the 3D. And 3D is way slower to set up. Lose hours in compositing and save days in modelling, lighting and shading. I think is worth it.

This tutorial will cover composite render out of Vray passes ( 3DS Max ) and comped back in Eyeon Fusion. 
 

Prologue.

Download this chart to visualize the next steps. You have three choices to get your render done. 

- You have no idea how to comp things togheter / your render sucks. Don't do composite render. Period. 

- You did a good job and your render needs no or little retouch. But you want to play it safe. Spend some time and split things down to a basic comp tree. No time at all? Just get the RGB split. Will be enough for colour grading and defocus.

- The project is worth it and massive post production can take place. Spend some serious time just to think it through. Drink coffee. Talk to the compositor (If any), ask what he might need. Visualize the next steps, what is likely to go under retouch and at what stage. Spend your hours to manually split the RGB file in components and accessories. Create output folders. Queue things up in the render farm.

You are going to do badass compositing here. If you are lucky. If not it will be way more badass. High end compositing is not a precise science. There are no routines and standard procedures. While few steps are quiet rigid and is just maths the major bit is made out of smart and personal decisions. Once you understand the compositing software and you understand the 3D project you will come up with the right thing to to.
 

Set the render passes up in Vray.

Do.
-In case you render HD, think doubling any pass in low res to make compositing quick and smooth. You will swap content in the very end.
-Plan to output any pass as single file/sequence in a single folder etc. Cramming anything into a container only begets a crazy file hard to manage, update, visualize. Is Not cool. Compositing is flexibility first.
-If you plan to retouch individually Direct light and GI consider ticking store with GI for any light meant diffuse ( Vraylights at the windows, etc). Else you get these 'diffuse' sources in the same pass as the sunlight. Which is just an example.
-Reflective passes alone will show way more noise than in the RGB. No solution but very precise noise treshold. That's the price.

Do Not.
-use any effect such as fog, dispersion, bloom. The 3D scene made with retouch in mind has to look quiet plain. Double check if the effect you think of can be achieved in post production instead.
-leave the VFB on. Render elements won't be saved on disk until the VFB is on. Do your previews and turn it off.
-use Vray Camera features. All the fancy shit in a Vray camera will wreck anything. No exposure, no vignette, no weird stuff. Even colour balance is lame. Standard camera is, in fact, gold.
-use color clamping. Over-range pixels are major bad! Learn to avoid it instead and look after your lights and shaders. 

Mind that some passes are children of the main RGB pass but extra stuff ( masks, AO, fx) comes from individually set up files. And these might look anything but the RGB. And have different content and even render engines. Muahaha xD
 

Basic Compositing.

Get hold on the pdf from above. The basic comp is quick to set up and gives you some room for retouch. You have enough passes to change, say, light temperature, hues, add a background, some defocus, bloom.
The output files are kept in simple TGAs  because we won't go mental, we like how the render looks already and retouch is more of a fancy.
The GI/Direct light coupling might look redundant but unfortunately the Total Lighting pass don't behave as expected so is not a simpler alternative.
You can see in the basic flow the only routinely nodes. There is only one way to chain in these passes (is maths), and once you've done you get a clone of the RGB pass. Ehm, the "Matching the RGB" thing is just a control tool to make sure the key passes of your render are correct. Is not the goal xD Once they match, move on.
 

Badass Compositing.

Ok, there is no good reference image here. Imagine, like... Hell. The shy flow from above grows in any direction, twists, fills the whole of your 30" monitor. And, to your despair, there is no 'right' way to do it, thus to teach. A few points, though:

-Increased colour depth. You now want some of your passes to store HDR data. Any pass involving Light gets promoted to 16bit floating EXR. Half float is enough and gives you tens on exposure stops. These passes can now be extensively fine tuned. You get über control over lights and reflections and the look of your scene can change dramatically. Mind colour clamping must be Off in Vray to output HDR data!

-Even more colour depth. ZDepth and Motion Vectors will greatly benefit from full 32 bit floating format, returning some striking results. Zdepth: do not use the render pass outta Max. It does not work for advanced defocus. God knows what is wrong with it. Use a custom black-fog scene instead. Motion Vectors: you know you have to find out the speed of the fasted motion in your scene, first. Plugins like ReelSmart and Lenscare will do the rest.

-Extra passes. Now that you Can control the image you want a fair load of both controller and content passes. Controller passes are Colour Correction Masks, Light Correction Masks, Normal Pass, Shadow Pass, Alpha. Content passes are alternative Reflection Pass, Diffuse Pass, accessory RGBs, FX. Any project will call for its own collection of content and controllers. No routines here.

-Render depth. Fusion uses the built-in bit depth of the footage you load, so you end up with a mix of depths ranging from 8 to 32 float bits. A quality flow has to be rendered in 16 bit, though, to preserve all of the subtle shades. Placing a 16 bit dummy background at the very root of your flow ensures you won't go accidentally lower. Downsampling to 8 bit the preview render in Fusion settings will keep the preview times down.
 

Aftermath.

That was, of course, just a quick overview of the compositing routine. A succesful render implies way more that what is here. I tried to put it in a simplified, logical scheme to give a bird's eye view of the insights. It is not a tutorial about Fusion, nor about Vray Render Elements. There are plenty of narrowed down tutorials on that, out there, but I always struggled to find a comprehensive one. Hope this one will take the niche :3