Havenhurst - Architectural Walkthrough Animation - 1080p Workflow Montage

[n8skow]

Thank you for that thorough reply.

Does checking the 'enable color management' box just take eliminate the adding a gamma correct node to every file texture? Or is there more going on behind the scenes?

[Asephei]

@n8skow Yes to your first question. To your second question: You check the box, and then all file nodes are assumed to be the "default input profile" (which is sRGB), and are remapped to the default output (which is linear sRGB). This takes care of most color textures, but you should set it to the same as the default output profile (so that no remapping takes place) on "scalar" or "data" type file nodes; displacements, bumps, normals, reflectivity, as well as for HDRs.

[Asephei]

@n8skow Basically all Color Management is doing is remapping file nodes, and the default setting takes care of all 8-bit per channel "color" textures, but it's important to not remap those scalar-type textures. Your textures have been remapped properly into linear space. Now you need to see the entire image in sRGB color space. Use the options for the Render View to do that: Render View > Display > Color Management.

[Asephei]

@n8skow Use the options in Render View for the display LUT (Image profile: linear sRGB, Display profile: sRGB), or use the tone mapper with gamma set to 2.2. If using the options in Render View, set gamma to 1.0 in the tone mapper. Basically you don't want to double up on the gamma correction (gamma at 2.2 is "99%" the same as sRGB color space). The only problem I have with the Color Management feature is the translation of textures for each render, besides that it's a good way to work.

[Asephei]

@n8skow I said "LUT" (Look Up Table) but to clarify, I'm meaning it in a sense that even a gamma correction is "like" a color look up, but LUT files have nothing to do with the typical sRGB linear workflow. You won't be using a LUT file unless you're in a production studio or certain project that requires it, as LUTs are very specific and might be used to calibrate the CG renders to match a custom color space, as a simple example. In Nuke you can create 3D LUT files if needed.

[n8skow]

You mentioned your 'previous' use of a linear workflow was incorrect, can you outline the steps you would use if you to redo this project today? Are you familiar with Maya 2012's current (broken) implementation?

[Asephei]

@n8skow I tested the Color Management (CM) in Maya 2011 (and 2012), and the HDR Render View with sRGB display LUT, under Render View > Display > Color Management, and all seems well in principle. The idea would be to use a tone mapper in the camera lens and set Gamma to 1.0, use the view LUT for sRGB display, and have each 8-bit per channel "color" file texture left alone, but set "linear sRGB" for "scalar" textures like bumps, normals, etc. HDRs set to linear too. However ...

[Asephei]

@n8skow The only reason I didn't use Color Management is because when it's enabled, it appears that mental ray has to do some sort of translation of all color managed textures on one core, separate from the geometry translation stage; each and every single time you press render, or IPR, you have to wait for all your textures to "convert". Even if they're MAP format. When you're just using gamma correct nodes and no Color Management enabled, this pre-render slowdown does not occur.

[Asephei]

@n8skow So here's my current method: On each 8-bit image, add a gamma correct node, and load my preset sRGB de-gamma (0.455,0.455,0.455). Leave all "scalar" (like bump, normals, specularity, etc) maps alone, same goes for HDRs. Add tone mapper (with default gamma 2.2). Set up lights, materials, test renders, etc. When ready to render for compositing, simply set gamma to 1.0, use EXR format and 16-bit half (or higher) primary framebuffer; this maintains linear workflow in compositing.

[Asephei]

@n8skow Once you have a middle grey reference in your scene (material with 0.18 diffuse), textures you paint in Mari, Photoshop, etc will render the way you'd expect them to. Everything starts looking less cartoony and more photorealistic; you get excellent and slightly flatter results in the render itself, with latitude for image editing, because the exposure and tone mapping are being done properly. I'm applying these concepts to my current project and the results are much better.

[Asephei]

@n8skow If using the mia_exposure_photographic, you should set Burn Highlights to 1.0 and Crush Blacks to 0.0 before doing the Gamma 1.0 render for compositing; this will essentially provide a true linear mapping of values (no tone mapping). Only do this if you want to re-tone map your primary framebuffer in the compositing phase. For mia_exposure_simple, set Compression to 0.0 and Gamma to 1.0 for an untone-mapped linear primary framebuffer render.

[Czarina999]

I'm speechless! Can you recommend training dvds or website? how long have you been studying

3d? I'm currently going through 3dbuzz's advanced modeling dvd. Do you have any tips or advice for someone who is just self-taught? How many hours a day do you practice when you were still a newbie? pls reply...

[Asephei]

@Czarina999 Well I've been off and on with CG as a hobby over several years, though now I'm serious about focusing on it. I do a lot of reading; I mostly learn through free tutorials all over the internet. The Maya help file is actually really good to understand the technical details, and it's also filled with useful overviews of modeling, animation, dynamics, etc. The DVDs by 3DBuzz are excellent and are great for getting started with Maya.

[Asephei]

@Czarina999 It appears that most people avoid the help file, but don't; it has excellent information. You'll also learn esoteric tricks that will help make you a power-user eventually. Oh, and use the selection-sensitive Shift+RMB and Ctrl+RMB marking menus while you model with polygons; they make the process much faster once you memorize the gestures. In time you'll be able to model with no interface or menus shown. (Ctrl+Spacebar to toggle interface show/hide).

[Asephei]

@Czarina999 As for overall concepts, there's plenty of misconceptions I had about rendering when doing the Havenhurst project. After finishing that project, I learned photography principles, by reading all the tutorials from the Cambridge in Colour website. My renders have improved significantly. For example, using the mib_blackbody input to the whitepoint in the mia_exposure_photographic lens shader will help you color balance as you work. Neutral color balance is very important.

[Asephei]

@Czarina999 Also, when tone mapping, the default settings of the mia_exposure_photographic aren't bad. I brought the highlights up to make the preview more linear (to match reality in a radiometric sense), but it's about what your eyes see, so you should leave "Burn Highlights" at 0.00 or very close to it. I did tone mapping wrong in the Havenhurst project and it compromised my ability to make realistic materials. You learn from mistakes and you get better with practice!

[Asephei]

@Czarina999 Also notice how my cursor (in the video) rarely clicks and drags the manipulators to move items around in X, Y, or Z. You can use the middle mouse button + gesturing in a direction to automatically constrain your movement to an axis. No other 3D software I know about has such a time-saving feature. Once you get used to it, you'll be able to position items fast, but I'd recommend a mouse with an easier to press middle button. I use the Logitech G700, which is really good.

[Asephei]

@Czarina999 As you initially light a scene, use a reference material that is 18% grey (which is 50% grey perceptually to our eyes), achieved by simply setting diffuse color value to 0.18. To see it as 50% perceptually, you need to have gamma correction of 2.2 (roughly sRGB) applied. Make sure to "de-gamma" your 8-bit color textures (not the scalar ones such as bump maps). Search the internet for "Understanding Gamma Correction", the tutorial by Cambridge in Colour explains it best.

[Asephei]

@Czarina999 When you render with an 8-bit per channel texture, linear lighting math is happening on gamma-encoded (non-linear) images, but the result "looks right" because the 3D software isn't applying a display gamma. It's highly deceptive. Use gamma correction and you'll know how a lambertian sphere actually looks, and quadratic falloff lights won't "blow out" as easily. This important concept is called "linear workflow" (Color Management really). Search: Maya linear workflow

[VEGETADTX]

AMAZING!!!!

After this I just want to ask you - do you hold some kind of courses for this?

I wish I was able to do this however:

1.I still don't have enough knowledge of MR.

2.My PC is getting old and pretty weak for this stuff

3.Mental Ray in Maya is way too unreliable and unstable, at least it has been for my PC so far(yet seems so much more realistic than Vray to me, regardless of what so many other say).

As much as this amazing video inspires me, it also makes me depressed :((((

[Asephei]

@VEGETADTX Thanks for the praise! No, I'm actually completely self-taught, I've always done things on my own initiative, no one "pushes" me, I have no mentor either; I read software manuals, look up tutorials online, etc. Occasionally I'll buy training DVDs, such as 3DBuzz's "Mastering Maya Advanced Modeling"... 116 hours of amazing training for only $100. Good stuff. 3D computer graphics has been a hobby for me but I'm working on going pro at it, lots to learn and improve on as always.

[Asephei]

@VEGETADTX I'd rather not get into debt to pursue my interests; I've never been to any higher education setting, I have no college time or degree in anything. As for the processing power, you should build your own system. Build a multi-CPU server for the best performance, or just build a single-CPU desktop, which costs less and is what I did. Takes a bit of research on computer hardware and building PCs but it's the best deal; don't buy a pre-built machine if you want the most out of your money.

[Asephei]

@VEGETADTX I tried VRay and felt it's just not nodal enough compared to Mental Ray; it's not a bad thing, I just like the big toolbox Mental Ray provides. I also like that you can write shaders for Mental Ray, so it has a lot of room for expansion. VRay has a few minor advantages, and it's a bit easier to learn. Read some of the tutorials/guides by mental images for Mental Ray; "Architectural Design", "Production Shader Library" pdfs, etc. They're free and you'll learn a lot.

[Asephei]

@VEGETADTX As for Mental Ray in Maya, with patience and practice you'll grow comfortable with it, despite the genuine problems of the implementation in Maya. Some problems people face are not because of bugs. For example, read the section in the Maya manual about "Contrast All Buffers" (in Render Settings) and you'll know why render passes are "slow". Add boolean attribute "miExportCCMesh" to Subdivision Approximation nodes to get seam-free displacement renders across UV tiles. Etc, etc.

[Asephei]

@VEGETADTX Oh, and check out the blacksmith texture set from the Foundry's website; free production data to demo Mari with. The whole set is comprised of 63 4096x4096 texture maps per channel (~1/2 the texture detail per channel used in each character model for the movie Avatar). Set it up with sub-surface scattering in Maya and watch VRay 2.0 take up more than 10 GB of RAM with memory mapped textures to render. MR does it with less than half that. It's the main reason I prefer Mental Ray.

[VEGETADTX]

@Asephei And I will definitely consider the Maya Modeling tutorial you've recommended, I feel pretty comfortable while modeling in Maya, but I'd say I'm still far away from kniowing all the modeling techniques you used in this video.

And what the hell are you saying! Hobby? With this kind of skill, and with just this SINGLE video as a portfolio, IMO you should get a job very soon, or at least be able to win a bid on the freelance project on some of the freelance sites out there.

[Asephei]

@VEGETADTX (1/2) I'll be a bit more specific; Zbrush smooths a surface differently than Maya does, so the base cage must be restored (or imported), saved as a morph target, then switched as displacements are calculated. Make sure to smooth UVs during export of maps; don't if you've smoothed UVs on dividing the mesh. In Maya, attach a Subdivision Approximation node, and add the boolean attribute "miExportCCMesh" to disable Catmull-Clark surface translation at render time.

[Asephei]

@VEGETADTX (2/2) Next, disable "Feature Displacement" (Maya settings) on the shape node. Convert the 32-bit (floating point) .tifs from Zbrush into Mental Ray's memory mapped format (.map). Use the command-line utility "imf_copy.exe" for this. Use a "for" loop in the command line to batch process an entire folder of images. Oh, and set the -p flag so imf_copy will store the .map files as filtered image pyramids. You'll then get correct displacement renders across multiple UV tiles.

[MrTheedarkhorse]

wans sure if it was real until the front door opened at 300 KPH ;]

[MrTheedarkhorse]

woooo0