Home Technical Talk

Workflow for building high numbers of buildings with damage states and no realtime AO.

zogthedoomed_UK
polycounter lvl 8
Offline / Send Message
zogthedoomed_UK polycounter lvl 8
i'm hoping for some pointers that might help me develop a pipeline for creating a large number of buildings for a turn based strategy game.

We're talking 80+ unique WW2 buildings from across Europe. So imagine cottages, town houses, factories, etc. with 3 states that show increasing levels of destruction at a poly and texture level. They have to be lodded and as the engine doesn't do any realtime AO ( no Unreal or Unity here ) I'd prefer to shy away from a construction kit approach. I'd want AO baked in as much as possible. So its nothing new and I'm sure its been solved manytimes before but I'm having trouble finding much information about it.

I suspect Houdini would be the tool of choice but I've only experience with Lightwave, Zbrush and Substance and have access to Max and Blender. While I can do most of what I expect would be required already, the scope of it is quite daunting so my needs are really about how to make it as efficient to produce as possible.

For example .. if I need to create a tiled roof, and then show it being gradually destroyed, presumably the tiles would be modelled and baked. So a high to low process would be needed. But I can't do that for every building, there's too many, so I was hoping for more a textural/procedural approach.  In this case I'm just about to investigate Substance Designers modelling and particle nodes that, from the adverts at least, look like they're exactly what I need. But then I immediatly have doubts about conistent UV's across multiple assets, their baked AO etc.. Well I've got to play with that a bit.

Any guidance, links or helpful hints would be gratefully recieved.

Thanks





Replies

  • gnoop
    Options
    Offline / Send Message
    gnoop sublime tool
    As far as I know you just model and texture sets of same buildings with different stages of destruction.  And AO  being baked in either separate unique UV channel or   vertex color or both depending on lods.    And then they just switch.    
  • zogthedoomed_UK
    Options
    Offline / Send Message
    zogthedoomed_UK polycounter lvl 8
    gnoop said:
    As far as I know you just model and texture sets of same buildings with different stages of destruction.  And AO  being baked in either separate unique UV channel or   vertex color or both depending on lods.    And then they just switch.    

    Thanks. I get whats required as a whole. But its efficiencies in dealing with the volume thats I'm finding tricky.

    Its an old engine and we'll need the usual colour, normal, spec, gloss so I'll be wanting to use substance for texturing them quickly and consistently. But that will require high res for baking each of 80 buildings and damage states and I can't see where the efficiencies can be made. I tackled a bigger job recently by leveraging Designer and Photoshop actions to do most of the grunt work which then left enough time to model and sculpt.  The damage states here require too much modeling, sculpting and baking, or use of dynamics, for me to do it in any reasonable amount of time. 



  • Mink
    Options
    Offline / Send Message
    Mink polycounter lvl 6
    Have you considered only doing bakes for a large, low res AO map, and leaving most of the damage states texturing, including AO, to be based on procedural texturing. Efficiency's gonna be the name of the game, and I assume you're not going to spend too much time careful inspecting your building models. Ambient occlusion maps rarely need to be perfect.

    Here's how I'd do it. Make medium detail models first, at 0 damage state, defer UV mapping, and go directly to making sets of interchangeable details using reused graph pieces in Substance Designer. Once you have a sizable material library, apply textures with vertex blending and loose UV rules for medium scale detail to UV0 in your modeling software. Then create your level 1 and 2 damage models, modify UV's as needed. Then, create a 0 overlap UV map for your models on UV1, and bake moderate resolution Ambient Occlusion map details. Blend large model specific with material AO in engine. Enjoy. There is absolutely no need to try to do things like normal bakes for simple buildings, especially when you're trying to make 80 of them on your own.

    Conserve resources, namely your time and the users' VRam.


  • Alex_J
    Options
    Offline / Send Message
    Alex_J grand marshal polycounter
    " There is absolutely no need to try to do things like normal bakes for simple buildings, especially when you're trying to make 80 of them on your own."

    I can second that. 

    Tiled roofs and stuff like that can probably come from materials - with generic stuff like this I doubt it's really necessary to be doing these things from scratch. If you are doing much baking and certainly any sculpting, I'd think you aren't making good use of plethora of existing resources for generic things like this. 

    It's probably too obvious to say but only assess the models in game as the player will see them to avoid wasting time on things that won't read. 

    I'd break things down first by material types and make a handful of generic modular pieces, then see how much you can create just with that handful. Fill in the remaining needs with more unique content only as necessary. 

    The project needs will change a million times for a million reasons and it will suck to have unique stuff get thrown away. So its really better to work with absolute minimum kit as long as possible IMO. Like one building only of each category. As the project goes on you'll keep refining your workflow and getting better idea about what works and what you actually need. 

  • zogthedoomed_UK
    Options
    Offline / Send Message
    zogthedoomed_UK polycounter lvl 8
    Alex_J and Mink, thank you. There's lots there to think about.

    Yes procedural texturing and building a library of materials would be the way I'd be going. With standardised ID colours I'd hope that side would be quick and relatively painless to iterate on.

    Our engine doesn't handle multiple UV maps unfortunately. And I'd love to have vertex blending, height based blending and a plethora of other nicities I had in Unreal. But we're still at a stage where I can request features so you never know.

    But your comments have suggested some other ideas that I'll explore. I agree that baking and sculpting is probably a step too far. I might still use Zbrush if only because I find it far easier to apply vertex colours for ID maps in that. I think with some clever use of filters in Designer I can get a lot of damage textures done fairly quickly. I'm curious to know how I can implement the modelling nodes for use in texturing or baking. It looks powerful but I need to get more confident with it.

    Thanks again for the suggestions and advice.

    Pat
Sign In or Register to comment.