Hello,
I’m having an issue sculpting my landscape with both splines and with the sculpting tools. It seems like my landscape has a very low vertex resolution or something. I am getting very jagged edges while trying to make small streams in the landscape.
I have made the largest landscape possible however I am running into this problem. My landscape settings are:
Is there something I am doing wrong?
Thanks,
RH
It’s 1 vertex per meter.
You probably aren’t doing anything “wrong” it’s just that the landscape system sucks.
You can try to scale the landscape down to get better resolution.
50 on X and Y will get you a vert every half meter.
So on.
The downside is you will reach a point where you compromise performance from having too many LOD0 components displayed at the same time.
And because some egged over at epic decided that distance wasn’t a good way to manage the LODs, you won’t be able to adjust at what distance precisely the LODs swap out.
If you aren’t looking to change things in engine, do yourself a favor and sculpt the landscape as a set of static meshes to bring in.
It’ll be much better all around.
The downside is that you have to use a different type of material to get physmat to work on it.
Not doing anything wrong. The landscape tool just doesn’t handle quick deep transitions from horizontal to vertical very well, as you can imagine if one vertex is very high and the next one very low you are going to see triangulation. You can work around by increasing the scale of your project assets (double scale of assets doubles resolution of the landscape mesh), avoid quick transitions from horizontal to vertical and small highly defined features, not just in unreal but in most projects you’ll see where they have vertical surfaces or sharp edges such as cliff edges sculpted landscape asset meshes are used. You get a feel pretty quickly what you can sculpt using the landscape and what you need environment assets for.
The landscape resolution is up to you really. There’s nothing wrong with scaling the landscape. The scale is what determines your desired landscape resolution, so if you want finer, like 25cm per landscape square, just scale down the landscape to 25%.
You have used the max landscape size possible, which amounts to extreme 8x8km. If you scale that down to 25% you will get 2x2km. At that point, performance won’t be great and you will have smaller landscape than you want.
The thing is, you need to decide if you really need as large world, or if you just randomly put in the highest number it took because you just wanted to have enough room. These are technical tradeoffs you always need to consider. If you really need large, 8x8km+ map for your game, and at the same time need less than 1m per landscape square resolution, at that point you probably want to use level streaming and world composition, and dynamically stream in multiple let’s say 1x1km landscape tiles. You don’t want to do that all in one giant landscape.
To be perfectly honest, I don’t really believe you need 8x8km landscape. If you are using in-engine sculpting tools to sculpt it, and you are sculpting so fine detail that you are running into resolution issues with the landscape, then there’s no chance you can ever sculpt 8x8km area in such fine detail in anything near reasonable time 
Thanks for the replies everyone! I’m relatively new to Unreal Engine (UE5 EA is mostly what I have been experimenting with), so I am still getting a feel for it. I am mostly using Unreal for animation and motion graphics, rather than game development, and it has been wonderful so far! Normally I use Cinema 4D, however this project definitely suits a (somewhat) real-time render engine as well as foliage etc. so C4D isn’t an option for this. That, and the fact C4D does not handle tons of polygons very well at all.
Actually I do need a very large environment. I should point out that this is project is for animation, and not for a game, so performance is actually secondary to how things look (unless we are also talking viewport performance slowing to a halt, which would of course not be good.)
I’m actually creating a real world area with lakes and rivers (using a heightmap) and then using landscape splines to create a very long trench in the landscape with raised dikes on each side as well as a bunch of roads. The real life area is over 1000km² but I’ve been reworking everything to fit into the largest landscape available in UE. The other challenge, is that it is flat prairie land, so I can’t use multiple meshes bashed together like the Nanite demos where it is all rocky, mountainous areas.
I think the best solution would be to scale up my height map, rather than scale down the landscape for more resolution. In either case, I lose a bunch of area. So my next question is, can I use multiple landscapes side by side? Like 3x3 landscapes forming a large square? And then splitting my heightmap in Photoshop for each landscape? Is this even a feasible approach?
I’ve been searching for quite a while for a 3D software that can use splines to sculpt trenches and roads, and that is how I came to use UE5’s landscape tools. The landscape splines are exactly what I need (other than being able to import splines - which would be great.) Does anyone know if there are other 3D sculpting apps that allow for sculpting with splines? I purchased World Machine and started to learn it, as you can actually import splines and such, but I feel I am still running into the low resolution vertex problem.
Ideally I would like to be able to sculpt a landscape in another software package using similar tools to the landscape system in UE, and then bring them in as Nanite meshes. However this brings in the problem of not being able to paint textures like you can with the landscape object in UE. I feel like every solution brings in more problems (just like almost every other animation project haha!)
Any suggestions would be greatly appreciated!
-RH
Not only feasible, but actually suggested. What you are describing is world composition and it’s a primary means of supporting large open world UE4 currently has:
That being said, UE5 actually expands this toolset further.
While this video doesn’t necessarily cover landscape specifically, it will still give you lots of tips you may find useful.
Thanks! I’ll look into this.
-RH
As I explained, there’s everything wrong with scaling the landscape. It throws off LOD distribution and degrades performance.
You can, it doesn’t mean you should
@RHMedia
Look into Houdini from SideFX
Their landscape tools work much better than the engine. You just have to learn, but they have tutorials.
Its probably best you bake it out as a mesh.
Particularly since you don’t need gameplay aspects to it.
Apply the foliage with the procedural foliage tools or manually instead of relying on the material…
Well, even the docs say that landscape scale is the primary means of letting you determine tradeoff between max landscape size and desired precision of your landscape topology.
The LOD distribution seems fine from my quick test, as it’s screen size based. Landscape at 1/10th of the original size draws about the same amount of triangles as when scaled to 100% size, and viewed from the exact same relative camera location.
The only performance degradation I can think of is if you reduce the landscape scale and then increase the landscape tile count to get back to the same size as original unscaled one. That way, you will generate much more geometry, so it’s obvious it is going to cost more performance.
That’s a misnomer.
The landscape LODs it’s components. Each component at LOD 0 is a drag on tris count. Particulalry when dealing with the largest sizes.
Mostly, the problem is the fact it is no longer distance based.
If it were left alone back in .18, then scaling up and down wouldn’t have a huge impact on tris count because you could just reduce at what distance from the player the transition occurred.
With the current LOD distribution instead, you’ll end up with multiple LOD0 on screen which will literally kill performance on some GFX.
I am having the same issues needing higher resolution ditches. What was your final resolution?
