PDA

View Full Version : Multi-pass rendering - worth it or not?



egearbox
03-27-2014, 07:41 PM
I'm considering, for the first time, getting into multi-pass rendering as I try to do some effects for a short film. I've done some single-pass compositing on previous shorts and I'd really like this next short to be a step up in quality. I know that multi-pass is The Way To Go if you're on a big Hollywood production with bunches of $$$ and animators and compositors but... it's just me on this shoot.

I'm wondering if multi-pass rendering and compositing (LW to After Effects) is going to be that much of a gain for me or if I'm setting myself up for a bunch of needless pain for very little gain? Should I just put the extra work into, I dunno - lighting? Color grading?

Anyone have any experiences (positive or negative) to share about this?

ernpchan
03-27-2014, 10:38 PM
Multipass rendering will help tremendously when you're in comp. It'll save you time from having to rerender out of LightWave.

erikals
03-28-2014, 01:43 AM
it can even save actual rendertime >


https://www.youtube.com/watch?v=kCJw2uDYpaE



also check Gerardo's links posted here >
http://forums.newtek.com/showthread.php?139419-Lightwave-render-The-Hague-Apartment

vector
03-28-2014, 02:20 AM
You should always render multi-pass. As ernpchan said,it can save re-renderings just adjusting layers/nodes. Obviously,you don't need all the render buffers. You must choose those you know you are going to need that aren't the same all the times

JohnMarchant
03-28-2014, 03:23 AM
I can recommend Janus for this. Takes a bit of setting up but is probably the best out there for LW. Lernie is always updating it and is helpful if you have any problems or issues.

gerardstrada
03-29-2014, 03:26 PM
As side note, render passes have been simplified a lot for studios that have adopted the physically plausible lighting and shading paradigm working together with real lighting contrast ratios and unclipped HDR-IBL solutions. Render times are commonly higher (due to raytracing), but multipass work is much simpler in compositing.

Also, if your short film is for you and not for a client, or if you are directing, then multipass rendering will be commonly simpler too or unnecessary depending on what you are reproducing in the scene. The opposite case is also certain mostly if multipass rendering helps to accomplish more quickly a very time-consuming effect straight from the renderer. Multipass rendering is there for helping. use it when that's the case is your choice.



Gerardo

geo_n
03-29-2014, 09:47 PM
For personal projects I only render out bg, moving items, effects separately with aovs that I would actually use.
With the new render engines coming out that are physically based, like octane, I would use less aovs since the render looks good already.
That's why I invested in octane waaaaay back even though I haven't used it in real work the potential to minimize time spent on render stuff and get a really good beauty pass right away is becoming closer. I assume octane will support more aovs in the future and comp guys can go crazy again if they want to.

Thomas Leitner
03-30-2014, 08:54 AM
....With the new render engines coming out that are physically based, like octane, I would use less aovs since the render looks good already....

An important proportion, that octane renderings looking good, comes from the internal tone mapping. But that is not very useful for compositing. Mostly it's about combining different image sources. Itīs therefore important that the tone mapping (and the like) just happens in compositing. Even if you can turn off tone mapping, unfortunately you still canīt export all needed buffers (but Iīm sure this will come).

Besides, it always depends on the needs. Sometimes a beauty pass (maybe with some alphas for separate corrections) could work. A depth buffer may be useful to adjust the depth of field in compositing (to save rendertimes, to make fast changes or in order to better adapt it to existing footage). The same applies to motion buffers for motion blur. Ultimately, the compositor should decide what he needs.


....I know that multi-pass is The Way To Go if you're on a big Hollywood production with bunches of $$$ and animators and compositors but... it's just me on this shoot....

So if you compose it by yourself, I think, you should expand your compositing skills too. If you donīt know how to use extra passes and handle them in compositing, they are worthless.

But hey, extra passes look good in the breakdown video ;)

ciao
Thomas

Greenlaw
03-30-2014, 10:08 AM
I do all my renders in multiple passes--it's standard procedure in any professional studio. You gain far more flexibility this way and you save a ton in render/re-render time (especially when it comes to motion blur.) Plus, you can take your shots far, far beyond the looks and effects of using LightWave alone.

My typical break down is:

'BG' (sky)
'Enviro' (terrain, general background)
'Char' (characters. Alternatively layers for other animated items like vehicles, fx (i.e., Hyper Voxels,) etc. In this case I use the name of the item like 'Car', 'Ship', 'MissileTrail', etc.)
'FG' (whatever needs to go in front.)

Depending on the complexity of the shot, I might break it down even further. Usually though, creating a RGBAM pass will cover all your selections for post image tweak. This is a pass were items are color coded 100% red, green, or blue with disabled alphas, and a fourth item set to full alpha but colored black. This pass gives you four masks with perfect anti-aliasing, and it renders fast because you can disable all your lights and raytracing features. Usually, I only need one or two RGBAM passes.

I always use exr so I can embed special buffers in the channels. The channels I find most useful are motion vectors, depth, normals and material/object ID. I primarily use Fusion for compositing and if I name the channels correctly, they appear automatically for tools like Vector Blur (for motion blur) and and Bitmap Masks (M/O ID for instant selection masks for any surface/objects combinations). The depth channel is good for fog and DOF effects. I try to keep the number of active channels down to what I know I will use to save disk space, but I usually have all of the above active since it doesn't cost anything extra time to render them. I prefer using exrTrader because it's really easy to setup various presets for different situations and compositing programs.

Because I embed motion vectors in the frame files, I never need to render motion blur in Lightwave. This is a significant savings in time and it's how most facilities cut down their render times. Fusion's Vector Blur sees this channel automatically but some compositing packages many need the data 'flattened' to a single image first. You can find info on this elsewhere in these forums. In LightWave, you can get motion vectors for almost any mesh. Motion Vectors also work for FiberFX and Instancer (this is huge because you couldn't do this with Sasquatch or HD Instance.)

M/O ID is almost as handy except, in the case of LightWave, you can't include a Coverage channel, which means you don't get anti-aliasing. To work around this, I simply add a tiny bit of blurring and it often works fine. If I need absolutely perfect AA though, I just create a mask layer (i.e., the RGBAM layer described above.)

The Normal buffer is good for boosting lighting--in some situations, you can practically re-light a scene in post this way.

By having the above, I very rarely have to render a scene more than once from LightWave, and if you have a template comp in place, compositing can go very quickly.

If I'm really pressed for time (which is most of the time,) I keep the layers to a minimum and depend on the M/O ID data and additional RGBAM passes as needed.

Anyway, I hope I didn't just overwhelm you with info. At the very minimum, breaking things down to BG, MG and FG will give you a lot of flexibility in comp.

G.

Greenlaw
03-30-2014, 10:13 AM
BTW, all of the the above can be done using native LightWave. (I use exrTrader but you can use LightWave's compositing buffers tool in it's place. I highly recommend exrTrader though--it's much easier to setup and use, and if you use Fusion, it works perfectly.)

If you want to reduce some of the work required in breaking down layers, take a look at Janus for LightWave.

Hope this info helps.

G.

geo_n
03-30-2014, 05:43 PM
An important proportion, that octane renderings looking good, comes from the internal tone mapping.


Yep even with limited capability octane looks hot.

https://www.youtube.com/watch?v=j0myKj4h4dk

spherical
03-30-2014, 10:49 PM
I do all my renders in multiple passes--it's standard procedure in any professional studio. You gain far more flexibility this way and you save a ton in render/re-render time (especially when it comes to motion blur.) Plus, you can take your shots far, far beyond the looks and effects of using LightWave alone.

My typical break down is:

{SNIP} (I'd quote the whole thing, as some people ALWAYS DO (really annoying), just because this information is so important to learn—NOT because I'm lazy and just hit Reply With Quote and go about my busy day.)

Anyway, I hope I didn't just overwhelm you with info. At the very minimum, breaking things down to BG, MG and FG will give you a lot of flexibility in comp.

OK, next time we come to Disneyland for some (always overdue) needed R&R, can I pay you for a teaching session? This is really valuable info and so much is truncated in a forum post. Still, it gives me a glimmer into that which is possible. Thanks for the mountain-top description. I'll try to expand upon it. (Overwhelming is good, sometimes.)

Ryan Roye
03-31-2014, 10:39 AM
I do render passes as well... started out doing it in Blender, but have since moved on to HitFilm (since it was temporarily made free in January). I use passport renewed which simplifies rendering good deal compared to Lightwave's native options.

The typical setup is as follows in my production pipeline:

1) Background + Shadows

2) Foreground/character objects

3) Effects layers (as needed). IE: AO, depth pass, isolated surface effects, etc.

You don't always need a ton of passes to get a greatly enhanced level of post production control... the end goal is to save time otherwise spent re-rendering stuff and to allow for effects/adjustments that aren't easy to do in 3D.

dwburman
04-02-2014, 12:48 AM
It might be worth pointing out that there are a few (at least) basic ways to do multipass rendering.

One way is to break out a scene by elements (for example the foreground, midground, & background mentioned above).

Another way is to break out a scene based on lights. Kat shows this technique in a tutorial at http://sdlightwave.com/tutorials.html

The third way is to save out the individual buffers (diffuse, raw color, reflection, etc or the newer diffuse shading, specular shading, etc) and combine those in composition to give you full control over how the image is constructed... I.e. if you think the object needs more reflection, you just turn it up-no need to rerender.

It sounds like most people just mix and match the methods based on the needs of the shot.

erikals
04-02-2014, 01:02 AM
some good notes there Burman... http://erikalstad.com/backup/misc.php_files/smile.gif