electric chair chamber - xsi, nuke. Personal project

[maze]

Hi folks,

First time for me posting here! I started using xsi around 1 year ago.. as I got hired on a xsi studio in Montreal,
so I had to learn it super fast paced on my own... while working... coming from using 3ds max where I was really comfortable...
and having to learn all the passes, partitions, overwrites, ice, etc... workflow from xsi seemed daunting to start of... but gradually
grew into me, to a point where I am amazed how cool this soft is. So ironically a year ago I though... yeah well I ll start using xsi and then go back to 3ds max
when I am doing personal stuff and really want to be productive... to... ..man, I can't believe I was "doing 3d" without the passes, partitions, ice...etc.. from xsi... haha

So well I decided a couple of months ago to start a personal project to improve my knowledge of xsi across different areas (mod, shading, lighting...rendering..sim)
And so I though to do a small cinematic based on a concept art I saw long time ago from an awesome artist named John Liberto. Also I started to learn nuke on my own so this look like a good project to practice both.

I like to wear many hats when it comes to 3d, but lately I've been trying to focus on textures/lookdev/comp...

I'll post my progress up to now... so you can see the progression...
here is an image that explains some of the thoughts I had while working on this project.

Edit: forgot to add the link to the concept
http://www.johnwallin.net/site/pix/ele.jpg
Moderator edit: our URL tags don't like anything not beginning with "www." - HB


and here are some renders to show where I am at the moment.






scene preview (havent modeled an "actual room" to allow the light from the hdr comme from the back)


Also here is a link to the layout and camera:
http://vimeo.com/39979287

Moderator edit: Embedded the vimeo, remove if you don't want it embedded - HB



I ll really appreciate any comments, thoughts or advice you want to share, and I ll try and post some updates as I go. I am doing this on my freetime so I am not on a deadline or something..

I ll like to improve the lighting and add some volumetrics, but I am not sure if thats feasible to do with the wom shader and ies profiles? rather how would you do that?

Also I must say that my computer is dying while rendering this...! scene is currently at 16mil polys... and mental ray isnt unfortuntely massive polygon count friendly...
At first I wanted to put some basic shading per material type, metal, wood...etc but mental ray would crash every time... So I could only add shading to the glass... and even though
could not push the settings high... (I am using raytracing and have the settings "optimize for heavy scenes") I am planning to upgrade my rig in the coming week with a gtx580 3G vram. That ll probably help, but until then maybe you have some tips about scene optimization you ll like to share

thanks in advanced for your time.
Regards


-Manu

[julius]

hi,
if you have crash problems due to memory limitations, you can try to take a look at standins.

[Maximus]

Very nice how its coming, i like it a lot. I suppose you have that polygon count due to zbrush/mudbox geometry right?
Sadly everytime i hit the same amount i had to struggle with mental ray, spending a lot of time optimizing and tryin to get things faster.
Are you using final gathering at all in those scenes?

[milanvasek]

hi.
reading your story reminds me of myself several years ago... I also switched from max to xsi when i went to a studio using it and was amazed by it ever since

image is looking really nice. great amount of detail everywhere in models. i can't wait to see it with textures and shaders. only the glass pieces in window look as they are smoothed somehow and it doesnt look right.
as julius suggested take a look at standins. it might help. also polygon reduction tool works really well in softimage (reducing polygons, but preserving details nicely). it might help a lot on those sculpted models. you could also render it in several layers, but that might be a bit tricky with all that glass

[maze]

Hey guys

@Julius, never heard about standins, but good to know I ll take a look, is it like proxies or something like that? thanks!

@Maximus, thanks man! well yeah, I chose to go with actual meshes rather than using displacements. Because I havent unwrap anything yet...
Although not all the meshes are decimated for many objets like metals and wood I try to use local sub div and add some detail without the need of zbrush, I did a very quick video to show what I mean... same applies to wood, even concrete etc... as long as is not really close to the camera.
[vim]http://vimeo.com/49124958[/vim] also yes I am using final gather because of the Sibl setup and hdr. I ve read somewhere something about a FG Shooter but hadn took a look again, are any of you guys using it? does it works well? I haven't create a master "abc.fg" map yet because I ve been adding stuff to the modeling constantly. How do you guys approach this? Do you wait until all your geo is final to create a .FG Map? I'll post my current settings:


@Milanvasek, yeah I feel a lot more focused on actually doing stuff rather than complaining on why things are not working as they are supposed to... 3ds max users will get this one..
thanks for your comments! the poly reduct tool works indeed pretty good, I havent done any "major" optimization, because I ve been telling to myself that there was still many things to model... and that I would adress that issue at the end... but looks like not, and now I have a pile of things to optimize, so it might be a good idea to stop for a while, optimize all the meshes and then further model some additional props... etc.
As for the glass! yeah... its funny you point that out, I also think it does look weird. Problem might be because I am using the mia_round corners to correct the crappy (meaning non beveled geo) comming from the rayfire solver. So to smooth out the borders I ve added the mia roundcorners... ideally I would bevel everypiece by hand, but the geo is so crappy that I would probably end up modeling each piece by hand... which I hope I won t have to do... !! I ll try lowering the "roundness" values, currently at "0.2"

Actually the glass is pretty much the only refractive material so far... and I ll keep it like that. Would it make sense to render it out on a pass of its own? if so how can I do that? I mean it wond reflect the other objects if I render the glass alone? Maybe I can trick this with a projection on nuke..?

Thanks for your time.
Cheers!


EDIT: @Mod, Thanks for helping me out with those links!

[Maximus]

Hey guys

@Julius, never heard about standins, but good to know I ll take a look, is it like proxies or something like that? thanks!

@Maximus, thanks man! well yeah, I chose to go with actual meshes rather than using displacements. Because I havent unwrap anything yet...
Although not all the meshes are decimated for many objets like metals and wood I try to use local sub div and add some detail without the need of zbrush, I did a very quick video to show what I mean... same applies to wood, even concrete etc... as long as is not really close to the camera.
[vim]http://vimeo.com/49124958[/vim] also yes I am using final gather because of the Sibl setup and hdr. I ve read somewhere something about a FG Shooter but hadn took a look again, are any of you guys using it? does it works well? I haven't create a master "abc.fg" map yet because I ve been adding stuff to the modeling constantly. How do you guys approach this? Do you wait until all your geo is final to create a .FG Map? I'll post my current settings:

@Milanvasek, yeah I feel a lot more focused on actually doing stuff rather than complaining on why things are not working as they are supposed to... 3ds max users will get this one..
thanks for your comments! the poly reduct tool works indeed pretty good, I havent done any "major" optimization, because I ve been telling to myself that there was still many things to model... and that I would adress that issue at the end... but looks like not, and now I have a pile of things to optimize, so it might be a good idea to stop for a while, optimize all the meshes and then further model some additional props... etc.
As for the glass! yeah... its funny you point that out, I also think it does look weird. Problem might be because I am using the mia_round corners to correct the crappy (meaning non beveled geo) comming from the rayfire solver. So to smooth out the borders I ve added the mia roundcorners... ideally I would bevel everypiece by hand, but the geo is so crappy that I would probably end up modeling each piece by hand... which I hope I won t have to do... !! I ll try lowering the "roundness" values, currently at "0.2"

Actually the glass is pretty much the only refractive material so far... and I ll keep it like that. Would it make sense to render it out on a pass of its own? if so how can I do that? I mean it wond reflect the other objects if I render the glass alone? Maybe I can trick this with a projection on nuke..?

Thanks for your time.
Cheers!


EDIT: @Mod, Thanks for helping me out with those links!

Yes, standin are like proxies can look them up in the help they are pretty easy to use, tho i see your geometry can be optimized a lot, lookin at the chair itself thats overkill
Regarding final gathering i think you have way high paramters, can try lower the Trace Depth in the FG tab, you have reflection at 10 refraction at 10 and combined at 20, might want to try 3-3-6, should be way faster to calculate FG and wont notice that much difference, the scene also is quite dark so its fine.

I take you will render out an animation right? In that case you might want to look at FGshooter but i'm not sure it works in softimage, Kizin tried and something went wrong if i'm not mistaken might wait for him to jump here and give advice

The glass looks a little weird to me, it looks like a coke glass displaced, what are the settings on your material? can help to make it look a little more natural.

Again i really like how its coming
good job.

[maze]

@Maximus, thanks again man! I ll look into those standins then.. mmm well yes the idea is to render out the shot of 600 frames, like on the video of the layout.
But I ll have to optimize a (hell) lot as I dont have a farm available atm and also I want no compromise on shading/lighting quality... I will like to add dof bokeh, chromatic aberration, lens distort... light particles...etc on nuke... so it ll be cool to have an actual sequence to play with rather than stills... !

About the chairs, haha...well yeah its definetly overkill! but as I have nothing 100% fix yet I wasnt sure how close I was to put the camera in regards to the chairs, so I did a zbrush pass overall...without decimation to keep the mesh clean. I ll keep a couple of them high res and the rest I ll polyreduct them!

here is the shader currently:


I am also attaching a test lut I did in PS and nuke. Ideally I would remove photoshop from the equation once I figure out how to grade properly in nuke.
https://dl.dropbox.com/u/4138955/drop3d ... estLut.rar

If you have any ideas about color correction, you are also welcome..

[Maximus]

I think your glass geometry is thin? so you dont have actual thickness, the roundcorner might give that look, can try to not use round corner and use the "thin walled" mode in the architectural material "additional options" tab. But just sayin, if that glass was the look you wanted to achieve its fine, i think its a bit odd

I'm currently doing a similar big project so its nice to watch what you come with
I'll eventually post it soon, scene is quite big so i get to find the mood to screenshot everything eheh.
Whats an average rendering time for those images you posted so far?

[maze]

Not the geo is not thin (meaning not single sided), its actually a thin box rectangle I pre-fractured with ray fire. But as the mesh is not clean, It looked hard to do a thin bevel to all the edges which such meshes, so I decided to try using the mia_round corners, but looks like its giving some weird results even at low settings... So at this point its a geo problem, I ll probably need a decent glass geometry to get good shading results... current mesh..


Might actually model the pieces by hand... although I really will like to avoid that!!
Have any of you know if DMM or Implosia have an option to keep quads? when fracturing stuff?

Hey good to know about your project! hope you get some time to post some stuff soon!

PS: I wanted to do some tests without the mia round corners but mental ray is on an acid trip today and crashes everytime... it sucks...
So I'm better of taking a bit of time to inform myself about standings and overall scene optimization before going further

rendering time with only 2 colored lamberts and 1 architectural for glass.... is at 3hrs!! using I7 930, quadro fx1800, 24G ram 1600...
I really need to lower that if I want to even think of shading something ...hehe

[Maximus]

I hear you believe me. Those times are quite normal to me, sadly AD did a great job messing up MR. MR doesnt handle that amount of geometry like Vray, 3 hours for a lambert render is a lot but i'm used :/
You could try 3delight, it comes free for 2 cores and 2 cores of 3delight are faster than 20 of Mental Ray for this scene probably

I would remove all the edges in the top and bottom of the pieces, (the slices just to say) and use the hard edge on the perimeter edge, then apply a bevel that way both the surfaces should be flat and you wouldnt get any issues.

Another method would be to go into the Geometry aproximation PPG and set the Discontinuity angle to 0, that way you would avoid ngons shading problems, and it'll render as flat surface. I use it quite a lot with glass.
Never had a single good result with mia round corner, i never use that, its just bad.

cant wait to see an update

[maze]

Hi thanks for the advice Maximus, well I actually had some time off yesterday and could do some improvements to geometry.
using the polyreduct all around I reduced the total polycount from 16mil to 10mil polys, yes, there's still a lot, but it still does feel a lot lighter.

Something really cool about the polygon reduction tool is that you can reduce polygons selectively, meaning in different percentages along the same mesh... thats awesome so I can keep density only where is needed. Also noticed that even a 50% total base reduction to start of, did not affect the mesh quality (detail sharpness) to any noticeable degree. So thats Cool!

Doing a test render I am now at 45+ mins per frame on a single Lambert and the basic glass shader, thats a couple hours less...!

-------

I hear you about 3Delight, Actually Ive had a couple months gig around 6 months ago where I could play a little with 3delight... (I was mostly texture painting and modeling...but still could play a little) I remember being able to use the default arch shader, so thats good, as I kinda know it well. I remember I did a house mostly in zb and it was rendering displacements like butter at imax res... was pretty cool to see actually! ...still though I am very very new to renderman engines as my experience its being mostly raytracers ....mental ray and lately a bit of arnold at my last gig... (but yeah I can't get to play with it at home unfortunately). I did some test with 3delight 2 core version last night... using only environment light (meaning an actual .hdr plugged to the envshader) I ve went quickly through the documentation... looks like you dont need, a blurred img for lighting... thats nice.

/note:I ll give myself some days to do some tests and decide if its worth the learning curve and render time or either I ll keep optimizing the scene with MR. As it was before. Again this is a breeze in xsi using passes for each renderer....

I'll attach a couple of tests I just did, and post some questions hopefully someone can give some advice. Test is done with no lights and only an hdr connected to the env shader.
So here I have some questions


1) From that first test I ll notice a difference in reflections when not using raytracing, and it does look better to me. Can someone care to explain me why? (primary rays reyes, ibl on, point based color bleed.)

2) I already had setup .ies profiles for the wom shader in Mr. Obviously wom shader wont work with 3delight and not even sure about .ies profiles. So any advice on photometric area lighting using 3delight? Is there any light shader for 3delight?

3) I am not sure why happens that the camera renders a different view. I saw the render output and it does render something different although the correct camera is assign to the pass. Is there somewhere I need to declare the camera to 3delight for it to use it? I notice the FOV is wrong.

4) How does 3delight deals with reflective objects, metals like. I saw that turning "raytracing off" does gives a nice result at least in the example I posted on a single ball. What is your approach to reflective metals from a shading perspective?

5) How good does 3delight manages geo count in comparison to displacements. Currently the scene is at 10mil polys.... Although if there was to be a significant improve on render time coming from displacement usage, I could reproject the detail on simpler meshes in zb and output .32 bit displacements..

Thanks again for your time guys!

Regards.

-Manu

[gustavoeb]

great project maze! looking forward for the final result...
I see you are playing with 3delight, cant help you much with that...

your MR setup seems pretty good though, but your ray depth is ridicously high... I understand you have glasses and all that, but do you need that much reflections? specially with glossy materials that can become a problem. also, have you given unified sampling a try? I recomend you do if you are on 2013 (it was slower on 2012), seems to work pretty good for me... (must have local samples down to very low values for best results).

[maze]

hey gustavo, Nice to meet a fellow south american using xsi as well!

I ve been a bit busy this couple days, but from the tests I could do using 3delight, I am getting the impression that there's a ton of power behind this
render engine, as I ve said, I used it before but very briefly on a gig, so its nice to get to play with the 2 cores version.

Although from my tests, I am at 30 mins render only with the .hdr and no lights. I think that I shoud've started by using displacements all around instead of actual geo
if I wanted to go 3delight route. Also as I am new to this engine I am having a bit of a hassle matching the lighting 100%. Mostly in regards to light falloff and distribution .ies).
Although I am pretty sure with some time it ll work. But I ve decided for this project to stick with mental ray.

I ll take a look at standins as was suggested by you guys.

@gustavo, I haven't used unified sampling, although I ve heard some nice comments about it. I am currently using 2012 at home, I have a student license in this computer so I might be able to do the upgrade if its worth, as I still have the access to updates.

hopefully that wont crash the wom archlight shader ?(...we never know)

I ll try to give some news soon, keep bringing ideas. I do appreciate your time guys.

à plus

[maze]

...did a quick test:



I modeled by hand beveling the edges, result is much better and I am not getting that weird rounded look..
its actually not as long as I thought modeling using a grid and shelling/beveling all the pieces at the same time.
I ll go ahead through all the pieces.

[Mathaeus]

So here I have some questions


1) From that first test I ll notice a difference in reflections when not using raytracing, and it does look better to me. Can someone care to explain me why? (primary rays reyes, ibl on, point based color bleed.)

2) I already had setup .ies profiles for the wom shader in Mr. Obviously wom shader wont work with 3delight and not even sure about .ies profiles. So any advice on photometric area lighting using 3delight? Is there any light shader for 3delight?

3) I am not sure why happens that the camera renders a different view. I saw the render output and it does render something different although the correct camera is assign to the pass. Is there somewhere I need to declare the camera to 3delight for it to use it? I notice the FOV is wrong.

4) How does 3delight deals with reflective objects, metals like. I saw that turning "raytracing off" does gives a nice result at least in the example I posted on a single ball. What is your approach to reflective metals from a shading perspective?

5) How good does 3delight manages geo count in comparison to displacements. Currently the scene is at 10mil polys.... Although if there was to be a significant improve on render time coming from displacement usage, I could reproject the detail on simpler meshes in zb and output .32 bit displacements..

Thanks again for your time guys!

Regards.

-Manu

Just a few ( very personal) answers about 3delight for SI
1 and 4: if you disable ray-tracing, it's just a 'reflection' from environment map, that's all. Probably you don't want this in closed scene, doesn't make sense, imho.
2: photometric lights, as well as physical sun, are not supported in plugin for SI. Physical sky *is* supported. About additional shaders, hardly you'll find something like photometric lights. But, you can find stuff like 'area shadow map' or like.
3. I don't know
5. Renderman compliant renderers, like a 3Delight, works differently than MR or V-Ray. All geometry is already tessellated according to screen resolution ( or something else). By 'shading rate' you control the level of detail - smaller is sharper and slower. However, contours are (almost) always tessellated properly. In short, you can have as many displaced objects as you wish at very low cost. As well as any other tiny stuff, hair for example.

MR Shaders are just emulation, maybe for having a smoother first experience. Under the hood, this is a very different engine, again. MIA material is just (almost) the same mix of components - diffuse, reflection and highlight and so. More an introduction, instead something where you'll expect the same experience.

Point-based GI and AO are (again) different methods than MR final gather or ray-traced AO. Let's say, they likes curved surfaces more than flat ones, they likes a lot of small details, like hair.

Ray-tracing is generally slower than in MR or V-Ray.

IMHO the best introduction could be to read the docs of standalone 3delight (they coming with plugin). It's not all technical lecture.

[maze]

Hi Mathaeus,

Sorry for the delay posting. Thanks for taking the time on those questions! That surely helps for future endeavors,
for now I ll stick with mental ray mostly for time reasons (haha yeah it does sound ironic...!). Although I don't have a deadline I ll like to start shading/texturing soon,
And probably it ll take me a bit to figure out 3delight as I want. It ll probably help to start a project from scratch using that engine.

Yeah, I hear you about the 3delight manual, thats actually where I went at first ;) really well explained. Although its not until you test stuff that you can be certain...
There s really good points on 3delight side and does look awesome, knowing you can get displacements, MB and DOF almost at no cost...
also hair as you point which I didn t know.

On the other side, theres the raytracing part that bugs me a bit for highly reflective surfaces, also light falloff and distribution. I ll try to take a look at the shader link you ve sent!

---

I ve been doing some tests with the broken glass geometry (very time consuming!), and although it might look simple I`ve encountered some issues I didn't thought of at the beginning but now that I am refining stuff it does looks off... and even if in the concept is working, I felt that it wasnt looking good in the 3d scene so I've changed the glass a bit. So I ll try to finish some more tests and post what I mean by that..

Cheers.