Where are these formations that you scan?
In the San Rafael Swell, in central Utah, USA. The closest name on the map is “Castle Valley” (But its not the Castle Valley just north of Moab.) The San Rafael is a 50 mile long zone that has been uplifted like a bubble over millions of years. The surface of the bubble has been eroded the whole time, exposing the ancient sandstones of different ages like the colored rings of a jawbreaker. In one direction you can walk along a cliff for 50 miles, while in another you enter entirely new rock types every half mile.
Whats the deal with the goats?
It turns out that large male goats make the best pack animals for hikers. They can carry 40-50lbs and can handle any terrain a human can, unlike a horse or mule. Llamas are too cantankerous and require a lead rope (Goats follow like dogs), and besides, I can load goats into my old SUV with no need for a trailer. Motorized vehicles must stay on established trails, so they can only get you part of the way, if the so-called “roads” around here even allow it.
What is the polycount like in this collection?
View the “inventory” video, which shows each item from all angles and the polycount for each LOD. Video cards these days are less worried about polycount than in the past, but still, different environment types may have different polycount targets. I suggest adjusting your main “LOD Bias” setting (Found in the Project Settings menu) or adjust LOD Group settings on individual prefabs. If you’re really worried about polycount, as in a VR game with huge visible distances, you may want to nuke LOD0 entirely.
Will the textures hog all of my video memory?
All of the textures are shipped at 4K. I consider the ambient occlusion, roughness, and such textures to be way less important in this case. I’ve set the import for those textures (the “mask” image in HDRP and Occlusion in Built-In-Pipeline) to 2K, but I feel you could go even lower without noticing a difference. Only the stone “plate” category really even needs a detailed roughness setting (rock tends to be rather uniform), and occlusion isn’t all that detailed as-rendred anyhow. To me the most important map is the normal map– so I’d shrink that down last. Remember the rock all has detail textures to enhance the surface regardless of the original bitmaps’ quality.
I’m having trouble baking the indirect lighting.
As of now, Enlighten is deprecated, so we’re forced to use Progressive. It works fine (but slow) in CPU mode but models that have LODs (meaning “all quality assets”) cause the GPU version to give blotchy results with glowing spots and black spots and low resolution. For now I personally use Bakery, the very nice GPU light-mapper from the asset store. Alas, it crashes when told to bake light probes in HDRP, so I do a quick pass with the broken Progressive-GPU to bake those, and then run Bakery on the main world. Sorry, its not the fault of the Castle Valley Collection– Its just the state of Unity at the moment.
What about adding rigid-body-physics and colliders?
Every item ships with complete, accurate colliders carefully matched to the rock’s shape. Doing this with auto-generated or arbitrary-shaped mesh-colliders would be very slow, so this set uses only simple convex mesh colliders. Many prefabs use several low-poly colliders at once to achieve the correct shape. Adding rigid-bodies is super easy, and all the colliders have the same ‘sandstone physics material’ on them. Not all of the prefabs are closed meshes, though, and would look stupid rolling around. The spire, sediement, rock, and plate categories are enclosed and can be used as flying debris and destroyed with mesh-shattering plugins if needed.
What are you using to create your photogrammetry assets?
If you’re interested in photogrammetry, I strongly recommend “Zephyr” by 3D-Flow. Its fast and affordable, and has a free version. I use a Nikon B700 for the photography for now, and various tripods and long sticks to get the camera in position. I use 3D Coat for retopology, UVs, and De-Lighting work. Max or Blender are great for the other work like cleaning up/merging scans or creating colliders. I use Xnormal for baking textures. Eventually I may make a video detailing the whole process.
How do you handle pixel density for large and small objects?
Pixel density decisions are made with the object’s purpose in mind. Of course the largest cliffs have lower pixels-per-inch than set-pieces that are meant to be seen up close. Around half of the prefabs share a material with one or more other prefabs in the same category. For instance, all 15 of the smaller rocks share a single 4K material, while around 6 boulders share one material, and the cliffs may not share with anything. Sharing materials demands special UV coordinates for the indirect lighting to avoid wasted space in the lighting textures.
When I bake indirect lighting in my scene, I generate too many light maps. What do?
Drop the “light-map resolution/ texel density per meter”. A value of 10 works fine for me, and lower still works. If you go too low, you can get bleed between UV islands, and too high looks the same as something more reasonable in the end. The trouble is that you’re probably not used to dealing with meshes of huge size. The cliffs are set to be 1/3 the normal light-map resolution by default, but you can go in and edit each mesh within the prefab to go lower if you need to. It all depends on how you’re using this collection. The “Light map size” value limits the max size of a lightmap regardless of how big it would be based on the texel density values. Dropping this prevents one item hogging all the space, but keeps Unity from batching as many objects onto the same map as it can. Its all about trade-offs.
Whats the difference between HDRP and Built-In-Pipeline versions?
HDRP and Built-In have entirely different texture map setups and materials. Built-in puts the smoothness data in the color-map’s alpha, while HDRP puts it in the mask-map’s alpha. Built-in uses a stand-alone occlusion texture, while HDRP puts it in the green channel of the mask-map. Auto-conversions don’t handle the detail maps correctly, and so to save you hours of ‘fixing’ just to upgrade to HDRP I’m putting out 2 versions of the Castle Valley Collection #1. HDRP also has a few decals, some “layered” materials, and some extra translucency data. If you change pipelines and need the set you didn’t buy, just let me know and we’ll figure it out.
The rocks in my scene kinda look like plastic.
PBR textures require environmental lighting data from reflection probes to work. If you don’t have probes, sometimes Unity will just use the sky texture. The problem is that most skies just draw more sky down below the horizon. This light is coming up and lighting your model in a very unrealistic way from all directions, causing the edges to tint the sky color even in dark areas. That’s the plastic look! Be sure to bake a few reflection probes and some light probes and suddenly everything looks like rock should look.
I need to make some special stone objects for my game that match this collection. How?
Contact me and I’ll gather up some of the original photos so you have something to work with when texturing your zbrush sculpts.