Modular Japanese House Content
August 1, 2018
Here a little recap of almost all the 3D assets done til now for the Japanese Modular House project.
About The Author
Anna
4 Comments
Inu Games
Game Dev Blog
Here a little recap of almost all the 3D assets done til now for the Japanese Modular House project.
Wow, looks awesome! Have you encountered any issues with Lightmass baking using modular assets, like weird behavior between adjacent pieces? Any tips?
Also, I’m guessing you retopologized the cloth meshes after exporting from Marvelous Designer?
Thanks.
We have encountered some issues with past modular projects, but not with this one. I think there are two reasons: first, Epic really improved a lot its lightmass baking algorithm, secondly the Japanese architecture alternating wood and paper most of the time naturally avoids those issues.
For Marvelous. I don’t retopologize my clothes in the 95% of cases. I don’t like retopologizing the clothes because I don’t want to lose the UVs generated by Marvelous. If your UVs are like the paper pattern you would make to create the cloth in real world, the fabric will look right on your model.
So I use a little trick: I simulate the cloth at high resolution, when I’m satisfied I freeze everything. Then, I export the high resolution model (that I will use for baking) , I lower the resolution (making sure that everything is frozen, I don’t want resimulate) and export that as low poly version. Then I still work a little bit the model in Maya, in case something should be corrected or I need a two side low poly model (Marvelous generates some incredible high poly connection between the two sides: I just get rid of those and replace with a simple bridge). In really few cases a repotopologize is still necessary for some parts of the cloth, like for the shirt in this post, but in most of the clothes I did, I didn’t need any.
Neat, thanks for the info! I recently decided to abandon VXGI and start using Lightmass for the first time, so now I’m worried about seams and proper lightmap UVs, but maybe those aren’t as awful now.
Re: two-sided cloth, are you making that dynamic in UE4 via Nvcloth or Apex? One issue I had is that Nvcloth only really works with single-plane cloth, which is distracting because it can make cloth look unrealistically paper-thin. I wonder if there are shading tricks that could be used to fake “thickness”?
For the lightmass: if you encounter any problem, you can always increase the resolution of your lightmap UVs and even substitute the auto generated lightmass UVs for a specific mesh with your own. You can look at Lightmass as a kind of art, it’s not always automatic to have it working properly from the start, but once you succeed to tune it, your scene looks really nicer than with movable lighting.
For the clothes, we didn’t use Nvcloth or Apex. We tried a little bit with the clothing integrated in Unreal, but we encountered also the problem you mention above. Also, the clothing integrated in Unreal doesn’t support morphs, so we experienced a real disaster with animations, as they hugely use corrective morphs. In the end, we gave up with that, at least for the moment. For the curtains (those are static meshes), I simply use a little code in the material that made them wave a little bit (the same is used for foliage in some Unreal example). For people cloths, we still don’t have a solution, we’re not working on that now, it will come back later. However, all the thickness you see in different screenshots I have on this blog, it’s really purely geometry, I just extruded or bridged some edges in the mesh; sometimes I’ve exaggerated it to give more the impression. I guess maybe also some fresnel added to the material can help for that, but still didn’t test it.