[Official] New Terrain System

wow, I can't believe I missed this thread.
It's maybe a bit late to add my opinion, but I'm gonna do it anyway.

Who are you?
I'm a professional 3D and Technical Artist, in the industry for over 5 years.

What kind of game are you trying to build or would like to build?
We're working on cycling simulation software which features large outdoor-scenery.

How does terrain fit into that?
Our scenes are vast in size and almost always outdoors with lots of nature. A proper terrain-system is vital for this.

What use-cases, features, workflows do you have or would like to see?
I'm mostly interest in a good vegetation&trees system. I don't really care much about waving in the wind and stuff like that though.
There are a lot of things I would like changed/added, but the following are most important to me:
- I would like to see this bug fixed: http://fogbugz.unity3d.com/default.asp?602612_12fehc9640vmud5n
As it is a huge problem.
- Trees need to be able to receive and cast shadows. (without this, trees that lie in the shadow of a mountain look like they give light)
- I've tried the tree creator tons of time, but I can't get a decent tree out of it, because of this I always model them by hand. So either this thing needs to be vastly improved, or all the features that are currently exclusive to treecreator-trees need to also be available for custom-mesh-trees.

 

Any news about what happened with the new terrain system? This thread started a year ago

 

It's in IR&D on the road map.

 

Thank you goat, I'm really interested in this topic.

 

What exactly are those APIs and where can i find documentation on them?

 

Ohh, this is cool. I'm just happened to have a terrain editor in my game. This will most likely make it feel a lot more snappy! Thanx : D

 

Ya I noticed those last time I looked in the docs and saw they were there.

Being that the overall revamp is probably a major project, one thing I'd love to see is just a better api wrapper for getting and setting values in the different maps.

For example I wrote my own wrapper methods that deal strictly with world coordinates and hides having to figure out what part of a map I need to fetch based on what I'm changing. So I basically have this:

TerrainInfo info GetInfoAt(Vector3 position, MapType mapType); // world coordinate

SetInfoAt(TerrainUpdate update);

TerrainInfo is an abstract class with a concrete implementation for each map type. Internally the wrappers handle all the coordinate conversion and accounting for resolution, etc..

TerrainUpdate is there because I wanted to treat TerrainInfo as immutable instead of just changing TerrainInfo and passing it back in.

Pretty much the entire Unity terrain api is one big exposure to the implementation details, which frankly I don't need or want to know about.

 

Forgot to mention, that wrapper also calculates the minimum amount of the actual map needed for an update. I have a method that takes a list of terrain updates which is the way I normally make changes, as most of the time an update spans multiple coordinates.

 

To Terrain System team, can we get an update about this, a blog post or a thread about the progress of the current terrain system Rnd? doesn't matter if it's good or bad, or maybe some stuff not gonna make it into the new terrain system, you know it's better rather than keep us in the dark.

 

I agree with @rea

 

My main issue with terrains is the workflow in general (it's a pain in the ass), as well as the requirement of multiple external tools and u3d assets.

Right now terrain workflow goes something like this:

1. World Machine to generate realistic looking heightmap, splatmap for terrain texturing, distribution map for trees / grass / objects, global normal map, global color map, etc.
2. Import this data into Unity, which isn't always straightforward. Unity can import heightmap, but need custom tools to apply the splatmap, and I'm not even sure how to import distribution maps without something like terrain composer or a custom tool. Pain points also include having to flip your heightmap and other maps to get it to line up with your terrain in unity.
3. At this point, you have imported your terrain, but you might want to change something. This is the pain in the ass - the slow iteration time, and using multiple external tools and unity assets. You have to go back and forth between world machine, unity, terrain composer. Your outputs from World Machine are kind of immutable. I don't like editing my terrain height once I've brought it into Unity, because then my in-game terrain is out of sync with my world machine terrain. Etc, etc, you get the idea. Right now the workflow in general is a pain in the ass.

IDEAL TERRAIN WORKFLOW
It would be ideal if unity had tools that would allow you work on terrains with an in-editor workflow that allows quick iteration times.

1. In-editor tools that allow you to both procedurally generate height data, and manually hand-paint height data.
2. In-editor tools that allow you to apply erosion and weathering to your terrain (preferrably with built-in texturing). See erosion brush on the asset store for an example.
3. In-editor tools that allow you to easily texture your terrain, based on height, slope, erosion, etc.
4. In-editor tools that assist in multi-terrain scenes, like some better way to work on multiple terrain tiles. Stitching edges of terrain tiles, etc.
5. Non-destructive workflow for adding roads, rivers, etc.

It would be REALLY COOL if there was something like a basic version of world machine that was built into unity, with tighter integration with unity for texturing, and placing trees / grass / objects.


In any case, please keep us posted because it's been a while and we would like to hear what your thoughts and plans are for the new terrain system.

 

We recently switched from Unity to Unreal, one of the reasons was the performance and the amount of extra-work you have to put into terrains in order to get a decent performance. Unity Terrains lack the following :

- Automatic heightmap import with splitting terrains into chunks based on the total size of the terrain.
- An underlying LoD for those chunks and it's terrain data. In Unreal you can see how it gets split and how chunks more far in the back have an automatic LoD system applied to it.
- Proper billboarding of trees. The current terrain system has a fixed sized atlas texture for the billboard, resulting in blurry trees, the more tree variations you add, the more blurry the trees get because of the fixed atlas size.
- A decent terrain shader that is an equivalent of the standart shader. The current shader is just ugly so you rely on assets like RTP to make it look at least half decent.
- This is part of the terrain system, the post effects need "post effect"-volumes, so an artist can also design enviroments properly.
- Proper tools for placing textures, e.g. automatic painting of texture based on weightmaps (see Unreal)

In short, if you want to implement a decent terrain system just do it how it's done Unreal 4.8

 

I'm not sure unreal's terrain is a poster child for good terrain. I'd rather have something much better as that comment smacks of "I'm new to unreal and haven't seen all the problems yet"

While it's editing tools are fun and it has a pretty decent starter shader, ultimately, it's still a horrific drain on resources and you end up with something far less impressive in practise, so this thread (to me) is a good way of seeking improvements beyond that.

 

i'm working with Unreal for a year now and reported 4 bugs in total. All 4 bug reports had been answered within 48 hours, 3 of those bugs had been fixed within 1-2 weeks. I'm sure i don't need to tell you how many bugs i reported in Unity which haven't been fixed or even answered. So yeah, from someone who only used Unreal for a year i can tell you i'm not looking back to Unity or missing anything - it's the opposite.

 

Yet here you are.

 

Well, it's sunday and raining outside ...

 

I think Unreal has passed Unity on terrains by a fair margin, especially when you consider having access to the source.

But, they both kind of suck compared to what you can get if you want to shell out several thousand for a commercial terrain engine. And if I was making a commercial game that needed AAA terrain, that's what I would do.

 

Which commercial terrain engine?

 

Probably VoxelFarm or Terrain Engine.

 

Now which ever game engine creates the first cubical supported terrain is gonna be myn

 

Hey there, I am actually in the process of doing just about everything on your list! I should have a useable beta up soon!

http://forum.unity3d.com/threads/sandbox-next-gen-terrain.358391/

 

A way to rotate and scale trees, grass and rocks procedurally.

 

Can I has your stuffz?

 

One thing I am curious about is the number of people who are creating huge open worlds that actually release a game and is it in the engines best financial interest to put a lot of effort into designing around those types of games. Don´t get me wrong, the game I am working on is open world and big, but I am constantly downgrading my vision not because of any limitations of the engine, but just because filling a huge world with interesting things is too time-consuming. Making a proceduraly generated huge world with the same 5 trees and wolf model repeating ad nausum is easy, but even creating something half the size of Skyrim and populating it with interesting things takes forever for a small team.

This is kinda just playing devil´s advocate for Unity, but I certainly understand why they devote so much effort recently to mobile and 2d, since that is probably where the majority of their money is coming from. The amount of us that actually have the time and energy to finish our big open-world is probably not that many people and a tiny fraction of their income.

That said. My three requests are a hole-system which I am pretty confident they will definitely be doing, better mesh-blending, and finally a built in stitch/cut system for terrains. For the stitch/cut, there are already pretty good tools on the asset store that do both, but I hate being that much at the mercy of a 3rd party asset that could be abandoned at any moment.

 

We are working on a flight sim which has world coverage. Most content is semi-procedural, generated from real world GIS data. Whatever terrain system Unity will come up with, it will probably be unsuitable for us. However, what we need is a flexible and fast API which allows a 3rd party terrain engine to use some serious LOD without making the GC throw up. That is currently not possible.

Creating such an API would allow the best of two worlds. It keeps Unity from investing too much in features few people will use while at the same time allowing other people to further extend the terrain system if they so wish.

 

that looks interesting, I will be watching your thread.

I'm still of the opinion that there is nothing that comes close to matching the results of world machine. I'm working on tools that will import lots of different data from world machine. Height, splatmaps, some RTP specific maps like global color and normals, and also import masks for tree/grass/detail mesh placement. My goal is basically to be able to make changes in world machine, alt tab over to unity, and within a few clicks have all of the new data imported into unity.

 

I've just picked up Gaia, and it looks like it will integrate well with a lot of third party tools like World Machine, while simplifying the overall workflow. Since it uses greyscale "stamps" to generate terrain, mask areas for texturing and planting, and other things, you can generate these stamp templates pretty much anywhere. One scenario would be to use World Machine to create high quality eroded terrain features, including texture and other maps, and save them all out as greyscale images. No need to figure out how to apply splatmaps, erosion maps, vegetation maps, detail maps, etc. Gaia provides a consistent interface for doing all of that. Plus, you can reuse these features, layer them, and apply them to entire terrains, or mix and match them as individual features of a larger terrain. It also leverages tools like Photoshop, which can be used to draw paths, roads, rivers, and other features, before applying them on top of an existing terrain. I've only scratched the surface so far, but it looks like a much simpler, more intuitive way to create terrains than anything I've tried previously. The other great thing is that the final output is a standard Unity terrain, that can still be modified and edited with other tools afterward. Until Unity has their own solution, this looks like a huge improvement over not only their existing terrain tools, but most of the third party ones I've worked with as well.

 

Yes, this is a big boost in terrain workflow making creation very fast and staying easy.
And once it is done you can still manually tweak terrain areas for your game purpose using Unity standard tools.

 

Just for posterity in case admin looks through forums about bugs, I believe that Unity removed atlasing for trees because it created a bug that made trees on terrain unusable (too slow even with one tree). I remember the Unity programmer mentioned that was the issue and i remember wondering if removing it had any consequences. Its interesting to read this here and see it did. Hopefully there will be a solution that doesn't regress to the performance before they removed it (a temporary solution i think but a necessary evil)

PS Gaia looks really cool. I use Terrain Composer and absolutely love it, but it will be interesting to compare Gaia with it (just bought it on sale - go quick)

 

Gaia is great - Unity should be looking at it for the new terrain system. Holes, mesh, multisplat shaders, tree/detail/grass LODding, and Gaia. That's all.

 

Kai ora

Who are you?

Hobbist, coming from Elder Scrolls modding etc

What kind of game are you trying to build or would like to build?

Roleplaying style of adventure something like Morrownind, Oblivion and Skyrim

How does terrain fit into that?

You enter into a world full of life full of mystic energe and life characters you have to interact with and explore,
places to find caves dungeons to delve. a place where these NPC's live, have home farms towns etc

What use-cases, features, workflows do you have or would like to see?

A chance of a texture add-on that does just Splatmaps, Flow Maps and a way to place a Full normal map to the terrain, a colour or texture could be picked to use the flow map. The Splatmap colour could be used along with height/slop filters in Software like Gaia for blending. The full normal could be selected and addjusted with a blend slider or the likes to blend the texture normals into the full normal.

I also would like to be able to do texture blending, not 4 textures sand, dirt, grass and rock
but blends of at least two textures say for example I have three layers I have sand, mud, dirt blending on a slat like mask basis or a noise mask to form the ground layer, then I have grass, moss and a rocky grass doing the same to form a grass/cover layer and like wise for rocks these blends can then be filtered applied on a height, slop and splatmap mask. maybe a node based material system like UE4 has but not limited to 4 texture by the time you add normals, Spec; AO etc. A mean to export this terrian into say World Machine and add erriosn etc exopt that and the required Splat Flow AO etc maps import it all back into Unity and plop a substance created in the like of Substance Designer (Not that I fully understand that yet) and get all the built in substance tweeks to fine turn your terrain texture

 

Take a look at "Lord of the Rings and "The Hobbit" we have it all in a samll (by most standards) Country.

 

I'd be interested in finding what the common features would be, that would satisfy MOST people, and still give enough freedom of customization to satisfy most of the rest. Obviously, not everyone will be happy.

Just some thoughts:

• I don't see a practical solution that handles both height maps AND voxels. They just seem like they are too different to effectively handle with one solution. I'm mainly going to just focus on a heightmap based solution for the other points.
• Seems like the system should handle "chunks" of terrain, stitching, and also handle it for LOD chunks as well.
• A nice, clean API for the shaders, materials, and textures should be there. It should handle the standard stuff...global color maps, etcv. The material(s) should be easily configurable, replaceable, etc.
• They need to make some serious changes to threading, and to allow for terrain streaming. Threaded updates to textures and such.
• Split OUT the vegetation and tree systems into separate systems that -can- use the terrain, so that they can be modified easily without affecting the terrain. I feel they are too intertwined now (this may be better with speedtree), but in the old system just removing a tree and its collider was painful. Threading would probably help.
• I personally feel like road and river tools and such are separate issues, that should not be a bundled part of the terrain system.

Anyway...just my $0.02. I'm sure Unity is aware of everyone's wish list, but I hope they just focus on making an awesome terrain system, that can be open and flexible enough to satisfy 90% of the projects out there (again, excluding voxel stuff), but with enough power to customize it to satisfy another 5-8% or so

 

I agree with your points especially about the vegetation system that should be splitted from terrain engine. it could be become new system for mass object placement/painter. which should work with terrain system or mesh.
Road and river could be part of decal system which optionally modify the terrain.
the only hard decision here is going to voxel or heightmap, this part kinda hard since voxel terrain not perform good in mobile (with large distance). Maybe this bit overkill ,like to support voxel and heightmap then split it as a mesh for final build.

 

Yes, this is a easy way to quickly detail terrain using stamps like Gaia.

The terrain would need a complete set of shaders like normal map, PBR, heightmap, pom, parallax and let us use the more appropriated depending on the terrain needs from realistic to cartoon to simple look.

 

I realize this is an old thread but it's the only one I can find that talks about a revamped terrain system. It's been five months since the last update on a thread started in 2014. Any updates at all about a possible beta where we might see a glimpse of the new system. Maybe they are waiting until Unity 6 for the big reveal?

 

As anyone don't know, after this 2014 annoucement , the best thing to do was to buy plugins on the Asset store, Unity priorities have changed a lot, Unreal 4 is playing on some different areas Unity don't see an interest challenging.
This Asset Store option is still the best today.

 

yeah this thread just went on without any outcome. Nobody seems to be working on anything related to the terrain at Unity for the past 4 years.

 

Exactly. It's a very good opportunity for someone to step up and deliver a quality terrain asset.

 

It's on the Roadmap under research near the bottom, but above visual scripting.

 

I am just guessing, but a new terrain system could be a major feature of Unity 6.0 ? It will be interesting to see how they handle the roadmap vs marketing moment with the 6.0 announcement. They will steal a lot of their own thunder if they are gradually announcing stuff on their roadmap. It could very well be that they are only going to really update the 5.4 stuff on the roadmap and leave the big 6.0 features for the marketing department.

 

It would be nice to know if the new terrain system is worth waiting for. And maybe some general info on the progress made during the last two years, as terrain "research" makes me picture this:

 

I would also love to hear about developments on the new Terrain system, but i'm guessing it won't come during Unity 5.x cycle. Probably a major feature for Unity 6?
Seeing as it is still in the research phase in the roadmap, i'm not seeing it being released anytime soon.

 

It's been on the research roadmap since Unity 2.x. So not very helpful to have it there.

 

There are tons already. The good ones at about $3000-$4000. But that's not the point. Terrain is and should be a core part of Unity since it's used by many many games. It's like saying that OpenGL doesn't work on Unity, so we should create an asset. Yeah, or a whole new game engine too! great.

Anyway, just really annoyed by the Terrain thing. I've been here for years and I've been through "polls" of what you want for the next Unity Terrain to have or the feedback/request part and voting system... several times already. With no outcome. And I'm talking since 2008 at least we have the same terrain system!

 

Sneak Peek please. . . some images

 

Thank you very much for the update Andrew!

The upcoming vegetation sounds very promising as currently, grass and detail meshes look and feel outdated. Are there plans to make it interactive - touch bending? And what about SpeedTree integration? I'll quote Greg Croft from the SpeedTree forums: "For interaction with branches in a tree or anything more advanced like that, you're getting into the realm of skeletal animation, which goes a bit beyond the current SpeedTree integration with Unity."

My second question would be concerning Multi-terrain editing/sculpting/painting, streaming terrain tiles, floating point imprecision with large worlds and performance - specifically GetHeights/SetHeights. While I'm assuming these issues are being worked on, it would be great to read some more information about this from the terrain team.

 

A few of the things he talked about are already available on the asset store. Of course it is always nice to see it as part of the main engine.

The part that blows me away is how much progress and design they already have accomplished without really communicating with us, which I thought was the whole point of the roadmap....so we could get an idea of what was coming, and maybe some hints of when. It is just strange to me that we get an info-bomb in a fairly dead thread that is many years old and the new roadmap info for terrain hasn´t changed in over 12 months.