Multiple questions for a worthy cause.

[SPS]

I've been putting off this daring request for some time, but I owe it to my project to ask.

I've been able to muddle through some decent architectural still for the last 10 years, but have seldom needed to animate anything. I have offered my time and efforts to an uncle of mine who is putting together a documentary of an historic black college that was razed years ago. I am in the process of working from old photos and surveys to recreate the campus. I'm not really getting paid for it, but I figured the experience would be worthwhile. So far it has been fun.

I'm approaching the final push, where my skill level will be surpassed. With the continued advancement of Vray, I've never found a good time to fiddle with new features. Always focusing on paying the bills. So I'd like to throw out what I will be working with and see if I can get a few shoves in the right direction. Not looking for any hand-holding here, just some shortcuts to what I need to explore.

I'll need to deliver at least one 720x480 segment containing a slow pan of 4 or 5 buildings and many TreeStorm instances. I'll probably displace the grass too. Something in the neighborhood of 10-12 seconds.

Things I don't know much about are:
Does anyone field render anymore?
Should I motion blur?
Should I texture bake?
Should I Proxy my trees? Animated?

I've been using Physical Cameras and VraySun for my stills and plan to use the workflow here as well.

I only have 2-3 machines to render frames, so I'll be looking for settings that are efficient, but not jittery. I've read some of what others have experienced with animating trees and displaced grass, so I need some proven methods of conquering those ill effects.

I've lurked around this forum and others for years now, so I know I'm in the right place. Thanks to those that offer their expertise.

[Neilg]

If the video is going on dvd, fields are still used - but we've actually had suprisingly good results not using them and putting them on in AE. This was a job that was rendered in progressive and then we thought we'd try putting on a dvd - I'd do some tests before your final render for this one.

Motion blur - if youve got the time to use it, it just makes stuff look better. Everything ive done in animation over the past 9 months has had motion blur on, i'm fully converted now.
You also cant get away with rendering sans fields without it, looks far too sharp.

Texture baking ive found to be a pointlessly time consuming effort - think it's only really used if you're putting together a real-time model of the area. I tried it for my first animation and later realised i'd have saved more time skipping the process alltogether.

If you have more than three, proxy your trees. Animating them never looks as impressive as you expect it to, hold off till you get chance to do some tests with that.

[MikeHampton]

I agree with cubiclegangster and would add a couple of things..

Does anyone field render anymore? Should I motion blur?

I've only ever done a couple of animations ever that were rendered to fields. For playback on an interlaced television.. rendering to fields and motion blur are going to give you the smoothest flicker free movement you can get.

However.. that being said.. IF you use the video for anythign else, like for playing a digital file on a computer montor (which is progressive scanned), the fields are really going to )($()# it up and you have to either render it again non-fields or go into some video editing/compositing packages and try to de-interlace it (which I have always found to be hit or miss). Also.. since VRay does not have a down and dirty post motion blur like the scan line does, rending motion blur in Vray is really going to up your render times.

Other than those couple of projects a long time ago.. I have not rendered to fields.. because most of our stuff is played back in multiple formats. The problem with an interlaced television is when it comes to having one pixel thin horizontal lines. When the TV scans the odd fields.. the line might not show up.. but when it goes back to scan the even fields.. the line shows up. This on off "flickering" of thin horizontal lines is the biggest problem. It's called "interlace flicker" or "line crawl". What I have done to reduce this is using motion blur and sometimes an additional 1 pixel vertical blur on the image can drastically reduce or eliminate this problem. However.. you may still have some materials in the scene that requre more. Like morter lines on bricks still might be crawling or flickering even after doing the above. To reduce/eliminate that, I end up using a local supersampeling on that material.

Now as I stated.. since Vray doesn't have a quick post motion blur like the scanline.. and since Vray motion blur takes so long to render.. I believe the best option is to render out a VrayVelocity render element and in a compositing package you can use this velocity to control the motion blur in post. I haven't had much luck in finding tutorials and such on how is the best way to do this, so I may not be doing it the best way. Also note that depending upon which software you use, you may need a 3rd party plug in to do this. It seems like combustion may be able to do this nativily.. while in After effects you need a plug in (Like Realsmart Motion Blur). Also note that you may be able to save these channels in an .rpf or .exr file format and use it that way. Again.. I haven't had much luck in finding workflows on the different methonds of how to do all this. Anyone know of any good tutorials on this topic?

And my experimentation with After Effects and RealSmart Motion Blur seems to indicate that it's still a somewhat "fake" motion blur (like an object in motion might be blurred.. but it's reflection in another object is not).. not necissarily perfectly accurate like calculating it at rendertime, but for most instances.. all you are looking for is smooth movement.. not physically accurate, thus doing it in post like this is much faster and gives you the control to change the amount of motion blur on the fly.

Should I texture bake?

Agree with cubiclegangster.. unless you are doing something in real time.. the amount of time you are going to spend doing that.. is going to be greater than the render time you would save by doing it.


>Should I Proxy my trees? Animated?

Depends upon how many trees you have and how many polys are they? Creating Vray proxies is so easy... you might as well go ahead and do it and get those trees out of your scene. Especially now that with SP2 you can create animated proxies. So you can animate and proxy the trees if you wish. I would recommend that the most bang for your buck is goign to be having maybe one or two trees closest to the camera very gently moving in the wind and don't worry about all the trees that are way in the background. But only do this if you have the time.

[SPS]

Thanks guys, for the tips. I just knew if I jumped in here, I might be on my way to making sense of some of this. I've jealously watched a few experts born on this forum over the years for having the courage to ask some questions. Maybe I'll be able to repay in kind some day.

I've been messing around with a "one tree, one building" sandbox environment to "kick the tires" on some of these settings.

I did a 120 frame test with these settings:
Medium-Animation IrrMap (defaults) HS:50 IS:20
Multiframe incremental (no precalculated vrmap)
Brute Force secondary (default settings)
Adaptive subdivision (default settings)
AA Filter : Soften (6)
Most of my bitmapped materials still have Pyramidal filtering
No displacement in scene
VRayHDRI skylight with a VRaySky as a background env. map
VRaySun with a .5 intensity (set low on the horizon)
VRayPhysicalCamera (movie) f14 ISO200 exposure and vignetting ON
I have not enabled Motion Blur or DOF yet for my test

I output to RLAs and import into AE. My gamma does not match and I am still getting some pixel crawl in the tree and in the window mullions. I'm hoping that Motion Blur will fix that but it's close enough to warrant checking a few other settings first. I like the look of the Soften filter for the purposes of portraying an historic campus at dusk. I would like to see some more warmth to the color mapping without over saturating the image. I don't want to rely on post too much here.

I'm also struggling with another theoretic fundamental that will likely bring some comments, and that is using a linear workflow method. My understanding of the many particulars is vague at best and further clouded by the many takes on the direction. Without throwing around specific terminology (which I am certain to misuse), I'd like to state my understanding of the method and see what others think.

I've read where others clarify that the LWF method most closely articulates accurate physical light and optical properties recently available to computing power for the desktop PC. I can appreciate the purist approach, especially given a task of matching real camera footage or building material color samples. In adopting VrayPhysicalCameras in my own stills, I have attempted to use this method and find that it works pretty well. Another reason I ventured down this road is that I remember some convincing demonstrations of getting the most out of your rendertimes by not blasting the scene with default lighting and then correcting using color mapping techniques. I may be way off here. There's also the whole gamma minefield when dealing with other outside applications.

The fine folks at Chaos have single-handedly revolutionized the architectural visualization industry. There should be a statue somewhere. In continually refining their product, it has been hard to maintain a reasonable grasp of the very best method while maintaining reasonable rendertimes. Look at any thread discussing settings of a particular feature and you will find many recipes that work well for their individual champion. My fear is that I have somehow trapped myself with each foot in a different boat by inadvertently marrying two techniques into one ugly hybrid.

MAN, I'm getting sidetracked here. Sorry. You can tell I don't get out often. "Put the lurker back in his cage!"

I'll close with this thought. While I look forward to my upcoming discovery of some of these settings based on some back-and-forth with experienced users, I am still somewhat representative of the mindset that wants to have the settings handed to them and to get on with the predictable results. I know that users that have taken the time to more fully understand the fine-tuning may tire of this constant whining for the "make pretty" technique. Look on CG Architect at some of the Chicken Littles of the ArchViz industry and it is not hard to see why some of the secrets of really outstanding results are kept close to the vest. So, to those of you out there that have reached a level of competency and continue to assist others on their way up, you are truly outstanding. The industry would not be the same without you.

[Da_elf]

i render for dvd at 24p

[MikeHampton]

If medium-animation presets are giving you a consistant flicker free GI solution.. then that's great. I find that I usually have to bump up to High-animation presest. Although admitedly I haven't had a chance to fully use the animation-prepass mode. I've yet to see if by blending multiple irradiance maps together weither or not that allows you to use lower settings and still get a flicker free GI.

But if render time is going to be a problem.. I would suggest pre-calculating the irradiance maps. That way you may only need to render them out ever 10, 20 or even 30th frame or more, depending upon the movement of your camera.

Also.. I pretty much use light cache for secondary bounces.. it's so much faster and wouldn't use brute force unless I was having some kind of problem. Even then, I would probably try detail enancement before I would go to all brute force.

I output to RLAs and import into AE.

I'm sure that's just some setting somewhere... sounds like maybe it's getting double gammed?

Look on CG Architect at some of the Chicken Littles of the ArchViz industry and it is not hard to see why some of the secrets of really outstanding results are kept close to the vest.

Well.. I know you said that this is just a test in a "sand box". And there's lots of little details and such that you could add to the building, but in my opinion in most architectural exterior visualizations.. the best "secret" is outstanding landscaping. I mean sure.. there are buildings that speak for themselves and it would simply be hard NOT to create a great rendering of them... but most "ordinary" buildings, what I notice between an "ok" or even "good" rendering and an "outstanding" rendering... mostly comes from the attention to detail in the landscaping and the surrounding environment. Realistic grass, trees, bushes, flowers, mulch, rocks, shrubs, maybe some ivy growing up the side of the building, etc. Not only that they look realistic, but are landscaped in a way that would be realistic. Basically all the stuff that I never have time to put that much detail into but I know it's lacking.

Then not only the landscaping.. other enviromental details... like having people that are appropriate to the scene and doing things that help tell the story or set the mood of what you are trying to show. I know I have used endlessly (and continue to do so) RPC people in my scenes. This is simply do to budget and timelines.. not that I want to use them. Although RPCs are ok and a down and dirty way to get people into not only a scene but an animation, 9 out of 10 times the people are not exactly appropriate to what the scene is and the mood you might be trying to get accross. I just don't have the time or budgets to green screen actors into my animations.

Being on a college campus.. you would most likely want college students (duh?), maybe a professor or two. And you would want them to be doing what college students do... walking to class, sitting under a tree in the grass reading a book, small groups of them here and there hanging out and talking, maybe someone running to catch up with someone else, maybe some students out on the lawn tossing a ball around or whatever. Obvisously also dressed in whatever timeframe (1900s?, 1920s?, etc) you are showing. But also appropriate to what mood are you trying to portray. The first nice day of spring (which is what I described above)? But could be somethign else. The excitment of the week before graduation? Homecomming week? A bleak cold winter day where everyone is just rushing from one building to another? A sad day in history of the start of world war II where everyone is a little worried and anxious?

Obviously if the attention to detail of the landscaping is no small task, having appropriate people (particularly in an animation) is a ginormous task.. especially when most of us are a sort of "jack of all trades" and have to produce animations from the start to finish with little or no help from others. Maybe you have a team of 2 or 3, but it's not like most of us have a team of 20 or 30 people or more that each specialize in different areas like modeling, lighting, animating, compositing, video editing, F/x and so on.

I know this is getting off the original topic of what setting to use for an animation.. but on the topic of what "secrets" there are to exterior visualizations.. I think that unless you have a really fantasical building to work with.. that the attention to detail of the environment around the building is the biggest "secret". The landscaping can make a good rendering a great one, and the appropriate people set in the stage and doing appropriate things for the mood you are trying to create can take it from a great rendering to an emotionally engaging one.

Sigh... now only if we had clients that would pay for these things..

[SPS]

Thanks Mike. I did run a 10 second test last night with a slow panning camera and the "high-animation" settings. I also switched to Adaptive DMC for the AA. I lost my roping edges on the windows and am pretty happy with the results so far. My tree shade still flickers or dances quite a bit. I don't have the grass displaced. Does bitmap filtering affect the object's receipt of shadow? I don't know much about the new shadow mapping feature of Vray. I'm wondering if I should lose the VraySun and go with something with more area flexibilty.

Rendertimes for my simple test are in the 2-3 minute range. I agree with everything you said on entourage elements. I would love to find the time to add the extra details you mention. Finding / creating good skinnable meshes of turn-of-the-century African-American college students could be a challenge though.