Terrain Painting
tracked
RestrainedRaptor Resident
LL recently released a teaser trailer for PBR + triplanar mapping on terrain: https://www.youtube.com/watch?v=Qmbws6tNwMw
However, the same old issues will presumably remain the same:
- You can only specify the low and high points for texture blending in each corner of the region.
- The viewer will then determine how the textures blend together.
- The blending is random, and done on the client side, so everyone sees a different terrain!
For those reasons, region owners have extremely limited control over their land, and that's why most people don't even bother using it, resorting to mesh instead. I have personally tried to make beaches in the past and found it absolutely impossible without setting the min blend heights very high, or setting all 4 textures to be the same.
Region owners need the ability to actually paint (like vertex painting) the terrain, in a similar way to the terraforming tools. 4 textures/materials is probably enough for most people, but more might be nice. This would allow the creation of designated themed spaces, beaches, paths, and more.
Photo Viewer
View photos in a modal
Log In

Spidey Linden
Merged in a post:
Terrain painting: region material palette with 4-per-patch selection
Megan Starchild
Glad to see terrain painting in progress. One request for its design: please don't lock regions to just 4 materials total.
Good-looking terrain needs more variety than 4: rock, a pebble/scree transition around the rock, riverbed, wet dirt along the banks, paths, grass, forest floor. With only 4 slots, creators will keep covering the ground with mesh and alpha decals, and that costs more performance than the terrain itself.
Proposal, borrowed from how World of Warcraft handles it:
- Each region gets a larger material palette (12–16 slots).
- The region is divided into patches (e.g. 16m or 32m).
- Each patch picks up to 4 materials from the palette, and the paintmap stores weights for those 4.
Why this works:
- Shader cost per pixel stays the same as today (still 4 layers blended per patch), so low-end hardware isn't hit any harder.
- The palette can live in a GPU texture array, so the extra memory is predictable and bounded.
- Because it's per region rather than per parcel, it avoids the texture-loading problems mentioned for parcel-level swatches.
- The editor can assign patch materials automatically: painting with a material that isn't in a patch's set yet adds it if there's a free slot.
WoW has used this exact approach (4 textures per ~33m chunk, drawn from a larger zone tileset) since 2004. It's a big part of why its terrain looks hand-crafted without heavy geometry.
Megan Starchild
marxman1313 Resident Sailor Vasiliev Very valid points and there's a few ways Linden Labs could get around this with parcels.
Switch to indexed splatting. This one would likely be the best option. Instead of each patch holding 4 slots, each paintmap texel stores the IDs of its top two materials plus a blend weight. Every texel is independent, so neighbors can never use up each other’s slots, and a parcel owner can use any material in the palette right up to their boundary. The shader only blends 2 materials per pixel, which is cheaper than today’s 4. The cost is extra shader work to smooth texel edges where the IDs differ. This is a known technique in other engines.
Keep patches, but add a mode switch. In region mode, the estate owner controls the palette and patches. In parcel mode, each parcel gets a small budget, say 4 materials from the palette, and patches that straddle a boundary split their slots between the parcels. That’s workable, but more complicated.
marxman1313 Resident
Megan Starchild That sounds much more viable.
ladiablesse100bc Resident
Something else to consider - height maps.
The idea is that certain strata of terrain act as transition zones from one type of terrain to another (sand to pebbles, pebbles to grass, grass to rock) so that if you assign specific terrains to specific heights, you get proper blending between the two. I believe this is how SL manages it now (but it could be done a lot more cleanly.)
Another consideration is not just 'how high' but 'how steep'. A gradual hill that is 20m tall can have grass. A spire that is 20m tall with a steep face is likely not going to have grass, instead it'll be rock or scree.
marxman1313 Resident
I agree with all of this except for it not being per parcel. You make good points but in SL it wouldn't go over as well with individual property owners/renters who have edit terrain rights. If it could be swapped between region and parcel though that might work best for land owners/renters although I worry it would end up like PBR where we have "both" with it being incomplete and bad on performance if someone isn't doing something properly and well optimized.
Sailor Vasiliev
I don't see why this couldn't work per parcel too. Parcel owners should be able to choose their own terrain materials from a larger region wide palette, with the system automatically managing the materials per patch. That would give landowners creative freedom without significantly impacting performance.
Megan Starchild
+1 on terrain painting. One design suggestion: rather than limiting a region to 4 materials total, give each region a larger palette (12–16) and let each 16–32m patch choose up to 4 from it, with the paintmap storing weights for that patch's 4.
The cost per pixel stays exactly where it is today (4 layers), while regions get enough variety for rock, scree, riverbanks, paths, and grass without falling back on mesh and alpha decals. Since it's one region-level palette, it also avoids the texture-loading issues mentioned for parcel-level swatches. This is how WoW does it: 4 textures per chunk, drawn from a larger zone tileset.
Dictatorshop Resident
I would expect that this would require much higher resolutions than we have available now before it is even an option.
RestrainedRaptor Resident
Dictatorshop Resident The vertex painting resolution wouldn't need to be linked to the resolution of the terrain textures since it's a region thing. Being able to paint it with a resolution of 0.5m or 1m would be a great start. If they supported terrain painting with the resolution of a 1024x1024 image, that would be 0.25m.

Spidey Linden
updated the status to
tracked
Issue tracked. We have no estimate when it may be implemented. Please see future updates here.
NeoBokrug Elytis
I made an old feature request that's pretty detailed, and applicable: https://github.com/secondlife/jira-archive/issues/8548
Eren Padar
I have often thought it a shame that I can't set my own textures on my parcel. In real life, people have different types of grass in their yards, farms have different types of crops, etc. To be forced to use whatever nasty texture that exists on a region is a shame. Currently the only way I know of to circumvent this is to create "land meshes" that fit the curvature of the land. That is both a major pain and consumes resources.
Load More
→