FXWARS! THE RACE: David Schoneveld

#41

Ok well this is not very similar to the animatic I showed earlier, Burke had to work in Maya on a mac 6.5 I was on 7 pc aaanyway so he worked independent of me with only my reference. We worked together (all his anim tho) to come up with this, I’m happy with it. Burke would keep tweaking it but I need to get some passes rendering. so that said this is what it’s going to look like as far as the animation goes. Tonight I’ll build back in the scenery trees and such. So check it out, final animatic… (6mbs)

www.agentfx.com/fxWars/crashScene_anim_B04.avi

He has more shots but I just won’t do them not with 2weeks to go. And I don’t think I’m sacrificing much, I like this so much more than my initial animation.

#42

Man, you’re right, that new animatic rocks. Looking forward to seeing the final product.

What are the tracks emitting from the camera, though? Is it supposed to be POV out the back of a car?

#43

Oh my that is extreamly cool. I love the snow being kicked up. You got some skill and style.

#44

WOW Mr. Schoneveld :eek: … you’re in this VFxWars to!.. OMG!! i have your gnomons Fluids DVD, your animatics looks amazing :thumbsup: i cant wait to see it in Full Render quality :D.

Good Luck.

#45

thanks guys, ah, but I did have a professional animator do that :D. So it damn well better look good, kidding, I became a team with Burke Roane from Luma, because the comments on my first animatic, made me realise that it would take too much of my time to animate this to the level I (and this forum) wanted, and Burke had basically 2 days of down time in which he could work on it.

What I wanted to get out of this wasn’t the animation end. I know that I can animate. I want to learn more about lighting, mental ray, and compositing. Right now I’m learning about optimising Mental ray for memory management so I can render more passes, and faster. I want to learn how to composite all the layers together, like ambitent occulsion, trick pass, reflection, showdow passes and such. I do it at work but normally I don’t have as many lighting passes and I don’t have to worry about memory managment and BSP stuff. Which is great to learn about huge speed increases! If you have time try to understand as much as you can from this…

http://www.jupiter-jazz.com/wordpress/wp-content/data/tr4kv2/html/chapter4-MEM.html

I’ve also learned more about the process of fx doing it all. Like story boarding fx. Great addition to my work flow.

So yeah that animation is all Burke. a little of my direction and referenced to what I already had, but all Burke. I should have gotten him earlier, eh?

#46

MentalRay is tricky, but I find the default settings that come with Maya and 3D Studio Max aren’t any good.

Here are my performance tips.

Samples per pixel:

In 3D Studio Max they rate this as percentages, and in Maya as samples. Either case the rule tents to be the same. Keep your sample range no more then 3 units aparts. The wider the range the long the render will take. I’m not sure why, but it could be that Mental Ray has to perform more samples to see if antialaising is happening when the range is wider. So in Maya something like 0 to 2 is fine, but -2 to 4 is bad. Better off with 2 to 4.

Jitter: Jittering in MentalRay doesn’t effect performance, but by default it’s turned off in Maya and 3D Studio Max. Turn it on.

Bucket Size: Always work with larger buckets. Something like 64 or 128 can improve performance. Specially if you are using satellites.

Scanline: Some people have told me that they get better performance out of MentalRay with the scanline rendering turned off. This might be true if you have a lot of reflections.

BSP: There is no question that the BSP is the best way to boost performance of MentalRay. Changing these settings directly effects the performance of calculating FG and reflections/refractions. The default BSP depth in 3D Studio Max is 40. When I render my highway with 40 it takes 2 minutes. If I change this value to 1,000 it takes 58 seconds. That is a huge performance boost. What I can figure is that out of the box products like Maya and 3D Studio Max have defaults for people with about 1Gig or less of memory. If you are like me and have lots to spear then boost those BSP settings. I also set the Size (or width) to 2 not 10.

On the other side of the coin. Decreasing these values slows down MentalRay but greatly frees up memory. If you find you are swapping while rendering then lower the BSP settings.

Trace Depth: I always set these values to 2. Anything more is for illustration.

Final Gather

The default method that MentalRay uses to compute the radius of FG rays is to take the size of your whole scene and divid it by 100. So ray width = scene width / 100.

The best tip I can give for controlling the performance of FG is to never allow it to calculate the pixel radius values. Always define the min/max range for FG rays. I usually work with the radius in pixels of 35 to 30. Keep your range narrow (again wide ranges in MentalRay don’t give you better performance).

The trace depth for FG should be reduce to 2 for everything. Unless you are rendering a visualization for print, or you have very rich secondary lighting.

Some have told me that setting the falloff range for FG rays can boost performance. It does seem to help, but not by much. It depends a lot on your scene.

The Best Tip for Mental Ray users

MentalRay has to convert all textures into it’s native MR texture format (or MAP). It often has to do this on the fly while setting up the scene. One of the problems is that MR will use default settings when doing the conversion.

If you convert your textures to MAP by hand, and turn on the mip-map feature. The texture file will be larger, but rendering will be faster. Since MR doesn’t have to perform as much texture anti-aliasing filtering. A mip-map texture is one that has been scaled by 2 several times, and MR can use the best sized texture that has been pre-filtered.

So take the time and convert your TGA, JPG and GIFs to MAP.

The results

By using the same final gathering settings. I was able to render the same image that took 2 minutes in just 1 minute, but making the correct changes to textures, BSP and filtering.

Don’t live with the defaults.

:thumbsup:

#47

thanks for that! You definitely inspired some tests, one of them, I found that adding jitter does create a render hit. 15 seconds to 30. My test was only a torus over a plane tho, with final gather on. Anyway, each thing that you say takes internalizing which is my goal, to see something know why I want to change it which way to try first etc. To understand the settings even better. Each time I do these projects I get better.
Each topic I’m trying to learn more about I do a good amount of reading on. There is SO much and when you think you’ve found at least all the topics there are more! I like reading what you wrote because it’s nice to see things explained again in different words. I feel like I have to read some things tons of times before I really get it. Really get the when, why, what’s of it all. Mostly it’s all just not that simple, each are work differently in different cases and respond to certain settings, all takes more experience. Thanks again, for your insights.

#48

Nice MR tips! I am trying to do more with MR at the moment. One more tip (paraphrased from John A Bell’s great book “3ds Max 6 Killer Tips”:

If you are losing fine detail like wires, antennas etc, you might want to check out the Contrast settings in the Sampling Quality rollout.

Contrast settings determins how to weight the samples per pixel values. If your fine scene details get chewed away in the rendering it’s usually not because your max sampling rates is too low, it’s because the contrast settings are too low.

To fix this, keep your existing samples per pixel min/max values within reasonable limits. Increase the contrast values slowly. This trigger the maximum sampling values earlier and should fill in the fine details. The Contrast Spatial settings are primarily used for still renders and the Temporal ones used for animations.

#49

Man, I’m kind of bummed, My wife reminded me that we’re going to NYC this weekend from friday morning to monday evening! ARGH! I’m so annoyed about that. I definitely have to loose the motion blur, which was so key to the look. And even with that gone I’m not sure I can finish. I have all the elements almost complete, I basically have to finish by thursday, maybe can tweak the comp monday night.
I’m debating, finish it the way I want, and take another week, or finish it with compromises to the quality and turn it in on time. I guess it comes down to what I’m in this for. The glory :wink: or to learn more. Hmmm. Does anyone know the smallest renderable size one can turn in? I was doing 640x400 but now I think maybe I can get a higher quality if I cut the rez down to like 400x220 or something like that with black bars on top and bottom. I wish I could keep motion blur it so important for the sprites, snow and debris. Even if I cut down the rez I have 3 nights and 3 days to get the rendering finished. Being gone for 2 of the 6 weekends has taken it’s toll on my entry. I’m sure I, like everyone here is expecting to do a lot the last weekend. Imagine that gone. Yes it’s possible but shoot.
Well I think I’m going to cut down the render size even more, still cut out the motion blur and try to finish by Thursday.

#50

What do you have to render?
How many frames and how long per frame?
What version of Maya are you running?

#51

hehe, we can see we re not alone( i assume many people are in the same way…)
for me i take 2 week in the middle(hic!), and now have sooo many problems to try to finish before the 22…

The glory :wink: or to learn more. Hmmm
that is the question! :smiley:

#52

with motion blur the car render is 40 minutes per frame for 200 frames, the background which is almost finished is 20 minutes per frame, but this is all w/o even doing an ambient occlusion, reflection, and shadow pass! Argh, there’s no way, Maya 7 PC. well I do have a single core 2.6 with only a gig of ram… the gig of ram is what’s killing me. I was going to get a new machine but some money I was waiting on is delayed.
My plan is to do one for the competition and another as the better version, the following week or two. I don’t mind, not having the coolest possible entry. It will be looking cool with a little more time. I don’t think the rendering is the only issue, if I had a render farm, then no problem, because I can work and render. whatever, maybe this competition will be extended (hint, hint, robert) by a week or two, just to bring an even higher level to it all. Kidding, ah sort of, anyway I’ll post a good play blast with debris and the crash online tonight.

#53

…now it’s tonight, check it out one particle pass, the red channel would be turned white in comp the green white, more fade and blured. The other is the crash debris anim test.

http://agentfx.com/fxWars/crashScene_gndSnow_B12.avi
http://agentfx.com/fxWars/crashScene_anim_B11.avi

I’m feeling like I’ll be able to finish this by thursday, at 640x400 no motion blur. So not a huge compromise but I’m sure everyone is going to make a compromise before the end. I think I’ll still work on this after the challenge, might give it a rest for a bit first.

#54

woaahh, looks awesome. the crash debris really adds to the already cool animation. Good luck on finishing in time.

#55

nice work there dave! just found this thread. I’ll have to keep checking back, it is looking really cool

#56

Really really cool David!

That looks so promising, nice work for the animation, congrats to your partner and to you! The camera movement is really sweet.

You did a good compromise i think, that does a good impression already, as you’ll just need to add motion blur later…

Cheers!

Xavier.

#57

That does look very cool. I would love to see a sample rendering with motion blur. :slight_smile: Just so we know what we’re missing.

#58

thanks guys, here’s a slap comp w/o the explosion… As you can see there are still some issues to work out. It looks ok, but MR in Maya does hide things that I key the visibility on, so you can see the front of the car still there after the crash. The snow is much to wispy, I imagined it to me more clumpy, also too much white, I’m going to get a sky back ground image. I was also going to add fog and falling snow to the scene, but with me going out of town… prolly not. So this is not far from the final product, I added a low res render frame of the explosion, we’ll see how that blends in tonight, with the rest of the scene.

(20m)
http://agentfx.com/fxWars/crashSceneComp_v02.mov

#59

The motion of the cars, the flip and snow look very good. Wish I could animate like that.

#60

final is on the first page.