davpe

hey, I can't help you getting the result with stylesheets+cryptomatte. you can, however, inspect the attached scene that's doing more-less the same thing in more simplistic way. steps I have done in the scene: 1 - build your groups 2 - convert them to primitive string attribute with Name SOP 3 - loop over the named primitives with For-Each loop (which is basically looping over your groups as the attribute is based on those) 4 - inside loop, generate random colors based on current iteration number, so you end up with random colors per group 5 - bind the color attribute into a shader that's applied to your geometry (or convert the color to some arbitrary attribute) and render that instead of cryptomatte it only have one caveat which may, or may not matter, depending on what you doing - if you've got two objects with the same group names, it may not result in the same colors (as the resulting color is not based on group name, but rather on loop iteration number) if you want to retain certain colors for certain group names over more objects, just use the solution i shared before, only you have to assign a color to each particular result. obvious disadvantage here is that with many groups it would be a lot of manual work to do. but if you only have a few groups, this is the most straightforward way i can think of. in the attached scene there are both approaches, you can swap the result of random and manual in the shader. if you need to do this across more shaders, you will have to copy that over to the other shaders as well - but that shouldn't be a big deal. stylesheets + cryptomatte probably provide more comprehensive solution but also the most technical, with cvex scripting involved groups_id.hiplc

check out the hip file with an example uvid.hiplc

hi, you can create arbitrary attribute on geometry, read it in shader (bind VOP) and bake it into a texture from there.

hi no. in order to match real camera framing you need to match focal length and aperture size as well. then you should get more less the same view. Houdini camera doesn't take into account lens distortion and that sort of things thou, so it's never gonna be 100% accurate. These minor optical properties are typically matched in comp stage (or you'd have to write a lens shader to do that for you)

well it's not a question of trust, but your answer already presumes that there will be some sort of difference between applications. do you trust Mantra more than Nuke? shouldn't that be the same?

don't take this for granted, but i think such thing could be simply a result of some sort of image processing deviation. i mean having Mantra compositing the beauty pass for you, and then taking decomposed layers and trying to put them together in a different software may not be 100% the same thing (for various reasons), even thou theoretically it is. it would be fun to try and put together the same thing in let's say Fusion and Houdini COPs - maybe you would end up with another two results with some tiny degree of difference?

as far as i know it has to be a point attribute. shader is not related to this.

hm... it certainly works for me. I remember thou, at some point this was broken and the solution was to force the console pop-up in houdini.env config file. i think the env variable for this is HOUDINI_WINDOW_CONSOLE = 1 i'm not on windows now so can't test it... check this docs page anyways: http://www.sidefx.com/docs/houdini/ref/env.html

actually, you can't read groups directly as i thought, but you can easily convert them to a string attribute and read that: groups.hiplc

you can work with groups in shader via inGroup VOP - boolean output indicates membership (which basically is a mask for that group)

it goes to console that will pop-up as a separate window. you have to enable "output to console" option on rop driver too. it's not enough to dial a number on verbose slider in reder statistics. annoying.

maybe check out this article for a few tips https://vfxbrain.wordpress.com/2018/12/04/using-orient-attribute-with-instances/

from my experience - it largely depends whether the studio you work for can (and want to) support procedural modelling workflows or not. in last 5 years or so i've done a fair amount of procedural modelling work on films and commercials, despite i'm mainly focused at lookdev/lighting. the reason for this is that if executed properly, it can be extremely efficient and a huge time saver. a couple recent projects i was working on, required square kilometers of CG city (for feature film). everything was decided to be modelled a textured procedurally, basically with only one person needed to do all the heavy lifting in Houdini, and maintaining alone the whole library of 100+ assets and it's lookdev/texturing. Any changes needed in modelling or shading were typically a matter of minutes, or hours if it was a substantial update. Another examples would be complicated industrial structures, or any big natural or manmade structures really, landscapes of any kind... you get the idea. For that kind of things, procedural approach is just awesome and there is potentially a big demand. Obviously, if you aim to fall into a general "modelling artist" box, you probably will do much better with Maya (even thou it's totally possible with Houdini too - it's just studios typically don't support it as a modelling platform). As others said, this area is already quite oversaturated and there are many excellent modellers out there. With procedural modelling on the other hand, there is too few people that have mastered it on good enough level, so it's not as wide-spread between big studios who still rather model stuff by hand as there is guaranteed constant stream of skillful people. Having said that, for me personally, procedural modelling and texturing was my most used skill over the last few years. I'm not sure how much of it is just a coincidence and how much it actually reflects a raising demand for this kind of skills. Right now I feel like I could solely focus on that and there would be plenty of work. if you wanna go that way thou, I'd recommend you to start actively seeking opportunities in this area (and build a solid showreel), and companies that are keen using Houdini, rather than hoping for being able to utilize your procedural modelling skills as a "modelling artist" in a random VFX studio. as for other of your questions: - Substance apps - definitely yes, it's very much in demand now - Mantra is great for procedural shading and rendering super heavy weird stuff. Also the best integration with all the Houdini features (obviously). If you like it, then it's good for you (I like it). - How often and what other skills to learn? All the time and whatever interests you. There is no limit to what your skill set and you will never know everything so don't worry and just keep going as far as it satisfies you - as for workflows, i don't have any fixed paths i always go. i tend to asses each project individually, thinking about what it needs, where i will likely need to do changes later, and what is safe to have more hard-coded. that's always different. then i maybe steal bits and pieces from another projects and put it together in a way that makes the most sense in that particual case. not mentioning that every project teaches me something new so i always try to improve and iterate my techniques, rather that using the same approach over and over again. cheers.

if you only want displacement and tangent normals (or other mesh features), just remove any shaders from your objects and it should work right away. if you need to bake the shader too, you need to use Bake Exports VOP in the shader that's applied to the hi-poly object. that will actually pipe the geometry features through the shader to the texturebake ROP.