That Cgtalk Lighting Challenge

[Jason]

I spent a couple of hours throwing stuff at it.

If anyone would like the scene to play with I'll attach it. (It's just the distributed scene conveniently split up for easy manipulation in Houdini)

Original thread on CGTalk

[GaijinYakuza]

I spent a couple of hours throwing stuff at it.

If anyone would like the scene to play with I'll attach it. (It's just the distributed scene conveniently split up for easy manipulation in Houdini)

Original thread on CGTalk

27939[/snapback]

I'd like to take a look.

[Jason]

I'd like to take a look.

27940[/snapback]

There is the scene without the textures as they're rather large. Tell me if it works at all.

glass_no_textures.zip

[altbighead]

There is the scene without the textures as they're rather large. Tell me if it works at all.

27941[/snapback]

nice.. I haven't looked at the file but I am having a hard time with raytracing .It is taking so long to render .. I am aware of optimizng mantra and setting up reflect/refract in obj setting but still it is taking quite a while.

[Jason]

nice.. I haven't looked at the file but I am having a hard time with raytracing .It is taking so long to render .. I am aware of optimizng mantra and setting up reflect/refract in obj setting but still it is taking quite a while.

27947[/snapback]

Yeah; this scene is not optimized at all. It's all brute force raytracing against true subd surfaces (except for a couple of the bottles). The scene is designed to hurt renderers. Caustics, DOF, tons of reflection and refraction bounces. Ouch. It's a good scene to show off fast raytracers

Visually, the glass shader really could use an "absorbtion" factor (where the thickness of the glass determines how filtered/tinted the light is.) Unfortunately Houdini is still in a roll-your-own mode wrt this stuff - the VEX Glass is the same one from Houdini 4.

[Mario Marengo]

Hey! Cool!

That's looking nice Jason. Good challenge.

It'd be great to see how people here can push Mantra to get a good result. Maybe branch your setup into a "Rendering Challenge" forum?

Unfortunately Houdini is still in a roll-your-own mode wrt this stuff - the VEX Glass is the same one from Houdini 4.

... :whistling:

[peliosis]

Don't you have a feeling that rendering such thing in mantra or optimizing it using reflection maps and tricks is sets us on "lost" position at the start?

I mean, mantra is good for a lot of things but not for fast raytracing of such scenes. The time spent on optimizing and setting things up is just too long, and I doubt anybody can achieve a result as clean as with mental ray or vray or hell knows what.

What is better in spending all day on optimizing and still being very slow than throwing some ready made materials and one light, checking three checkboxes and having work done fast and beautiful in an hour.

Correct me if I'm wrong but I do still oppose the fight against unavoidable, and opt for MR in houdini:)

Not that I don't like your rendering Jason because I like it very much:), especially after some hard months spent trying to get the beast to work the way I like.

[Jason]

Correct me if I'm wrong but I do still oppose the fight against unavoidable, and opt for MR in houdini:)

Can you post some results for speed comparisions for identical scenes (and shaders) between Mental Ray and Mantra? I'm curious about the speed factors involved.

Optimization is extraordinarily necessary in feature production, by the way. If you can get something to render in 20mins versus 40mins, you get twice as much done. I've seen people bank on "quick" renderers (like VRay and Brazil) and end up with 22hr renders at HD resolution because their light half-video resolution renders clocked in at 20 mins. 22hrs at HD = ~35 hours at 2k. On the other hand, I've seen renders for I,Robot drop to 25mins/frame from 4hrs at 2k resolution using Renderman purely because you CAN optimize it. Only hobbysists and very turnaround commercials artists have any excuse not to learn how to optimize renders.

[GaijinYakuza]

Can you post some results for speed comparisions for identical scenes (and shaders) between Mental Ray and Mantra? I'm curious about the speed factors involved.

Optimization is extraordinarily necessary in feature production, by the way. If you can get something to render in 20mins versus 40mins, you get twice as much done. I've seen people bank on "quick" renderers (like VRay and Brazil) and end up with 22hr renders at HD resolution because their light half-video resolution renders clocked in at 20 mins. 22hrs at HD = ~35 hours at 2k. On the other hand, I've seen renders for I,Robot drop to 25mins/frame from 4hrs at 2k resolution using Renderman purely because you CAN optimize it. Only hobbysists and very turnaround commercials artists have any excuse not to learn how to optimize renders.

27953[/snapback]

Nice render and I like the way you set things up. I tried to render on my machine just to see what I could get, even with out the texture maps. It was very slow as you said. Just out of curiosity, how long did your render take?

[Rafal123]

beautifull render

[Mario Marengo]

Optimization is extraordinarily necessary in feature production, by the way.

[...] I've seen renders for I,Robot drop to 25mins/frame from 4hrs at 2k resolution using Renderman purely because you CAN optimize it. Only hobbysists and very turnaround commercials artists have any excuse not to learn how to optimize renders.

Yup. Print that and frame it!

Optimization (and learning how to do it) is not an option.

... must... optimize... must... optimize...

[peliosis]

Yup. Print that and frame it!

Optimization (and learning how to do it) is not an option.

... must... optimize... must... optimize... 

27956[/snapback]

I absolutely agree, BUT...

first: I'm rather not talking aboust big production houses like DD because the thing looks quite different since you probably can sit a few superbly well educated guys and tell them to optimize the scene.

In a small, say 10 person studio filled mostly with allround artists you just CAN'T do it.

secondly: With mental ray rendering for example gi or ambient occlusion, I'm 10 times faster and clearer from the beginning.

Vray from the box is yet much faster than mray which is just a plain fact.

Again I know about the wise men with long beards who can write their own shaders and whatever but this is absolutely another level of software usage.

I work with a few friends and we could never survive with mantra only.

If we have some work to be done including heavy raytracing we could choose to do it in houdini and spend days on optimization, trying to achieve something we already have straight after installing e.g. xsi.

(And I believe houdini can be a lot better with mr than xsi is)

Still some features like motion blur and depth of field which are a strong point of rmands most of the time can be avoided. (e.g. Is Blur using it a great deal?)

I'm not a techy guy and I will never be writing professional shaders and fiddling with code, so that kind of optimization doesn't interest me and many alike.

I would like to have a robust software to use in all cases to avoid switching.

What if you have to render the fruit basket from former challenge, 360 orbit, 400 frames? It's not a problem to do it beautifully with any of the loud raytracers, put shaders check five settings and go, in mantra..."ouch, where was this great Mario's SSS shader???, oh my now I have to make point cloud for every fruit. Glass, vex glass is freaked so...perhaps I should make myself one. Caustics, fg multibounce...ouch"

I'm sure all of guys not "equipped" with some dev team will second this opinion.

I believe these are some immportant things.

And again, I bow humbly towards all feature production studios but this is not the case. We here are small and are not going to be even quarter as big so some ready mades are really helpful.

[Mario Marengo]

I absolutely agree, BUT...

27959[/snapback]

Hey Peliosis,

I totally understand where you're coming from and agree with the fact that "the best tool" is simply the one that solves your problem. No one can argue against that. Full stop.

It's just a difference in the types of problems that need to be solved. In Jason's case, and to a surprisingly large degree in Axyz's case as well, flexibility is king -- if you can't get in there and tweak every aspect of your rendering pipeline and explode it into a hundred layers, then you'll likely ditch the renderer. But that flexibility, as you know, comes at the cost of a much steeper learning curve -- just the price you pay (unless you like the subject in which case it's no price at all, of course).

Having said that, no one could possibly disagree with the fact that, at this time, there is a severe lack of ready-made industrial-strength shaders with out-of-the-box Mantra. Sad but true. Sad because it doesn't help to show off what in my opinion is a very good renderer -- but this is not about plugging Mantra, I feel the same way about other renderers.

Just to be clear then, the point is that there *is* value in trying to optimize a scene like this one for a renderer such as Mantra or Renderman, and that taking an example such as this would be a good stress test for both of those. In other words, the fact that you can probably achieve a result quicker with a more specialized renderer (which is valid in and of itself) doesn't make the exercise pointless.

Cheers!

[Jason]

Oh, I feel your pain, indeed. Mantra (and in many ways, Houdini itself) has never been a slam dunk setup for smaller shops and I don't think there is a good excuse for it. Even the bigger studios usually only have a one or two good theoretical shader people. People like Mario here on odforce are extremely extremely rare.

But on the other hand, I believe Mantra has all the speed and technology it needs to satisfy 95% of all rendering needs and desires, bar a few truly fun and useful high quality shaders such as a good glass shader, car paint, raw metal, subsurface and so on. These "primary" surface types need to be implemented in a good solid way in order to tantalize people from smaller studios into using it.

[AlexPaes]

Hi,

Just arriving at the houdini world through the apprentice edition and i'm really enjoying it, i too feel the lack of nice builtin-shaders to make it easy to come up with nice good-looking scenes easier/faster.

I'd think the biggest problem with this is that houdini is quite film-production oriented and as such Sesi dedicate their precious time into developing what the big houses in the industry ask for. I think that if for smaller studios and specially for people trying to learn an app this might be a bit disapointing it also translates into the best professional-level technical support available to any of the 3d apps out there on the market.

Well, this is just the 2 cents from an H. Apprentice.

Cheers

[sibarrick]

100% agreement here. Personally I think they are missing a very simple trick here, and was very glad to see that they didn't fall into the same trap with DOPs, they put just enough pre-rolled stuff in that you can easily put together the quick and dirty stuff and then once you're hooked you're off and running. I sometimes wonder if they were really serious about getting more small studios interested in mantra it wouldn't be worth their while saying if you buy Houdini will write you half a dozen core shaders but at the same time we'll keep copyright on them and release them in the package.

Just my 2 cents, still keeping my fingers crossed for more packaged shaders in the next release. I really thought it would have been a perfect addition to the H8.1 release but it was not to be.

In small shops even if you know how to write shaders it still doesn't mean you have time to write them.....

[peliosis]

Thanks for a kind treatment after my long rant:)

I really hope SESI is reading such threads and will have a look on the subject, In my case rendering capabilities are more important than character improvements.

I wonder what troubles do they have with including MR? This would be a fast bullet proof solution for "small shops", it's already included in major packages. We could make GI, and other heavily raytraced passes in MR and the rest in mantra.

I hope they'll put more effort into rendering and I think this could be the key to attract many new eyes to houdini.

(pssst...they could also speed up the edit sop :whistling: )

[SvenP]

So many people are crying about missing default shaders in houdini so why dont we do something ourselves ? Maybe adding a shader library page to the wiki where everyone lists the shaders he wants to see in the library and those in here who know how to write shaders could pick a requested shader and implement it ( preferably compiled as vop so that the end user stays flexible ) . This wont be a thing that gets done in a day or two but if there are only 5 users contributing and everyone writes 1 shader a month we will have a nice library in a half year.I am shure i would find the time to write the one or other not to complex shader ( The really hardcore ones would be for Mario ) .

Sven

[DaJuice]

I like this idea Sven.

btw, anyone seen this thread? It's huge! A community effort at creating mental ray shaders... some really nice looking stuff there.

http://forums.cgsociety.org/showthread.php?t=104578

[Hwee]

That's a long thread. But the "community effort" looks more like one guy with a lot of cheering on going around him.