[Official] New Terrain System

Pretty Sure I saw you on the thread but

SEGI

It is still in beta and only has one developer but so far it is pretty awesome. The only GI system I dare to use in the current state of Unity. Especially on a terrain that uses trees since Enlighten is not very useful there. There are some improvements on the near horizon that will help even more with use on a landscape.

 

Interesting. I actually think GI is the only way to go for outdoors. The current baked lighting system doesn't work properly with shadows and without GI you don't get the nice atmosphere you need for an outdoor environment, especially when you combine with tools like uSkyPro and Prism.

 

Really disagree here because lighting may be more subtle in an outdoor environment but it has a major impact with making an okay scene or an amazing scene. SSAO cannot cover the bases when it comes to making decent shadowed trees. So while bounced lighting may not be as prominent in a more ambient lit environment there needs to be some sort of skylight occlusion.

Also there is the issue that if the option were ever available in Unity to have a large dense forest it would not look great using one ambient value unless the entire scene was only a dense forest. Bounced Lighting and Skylight Occlusion play a major factor here and requires the trees actually contribute to the lighting.

 

I agree completely. I'm also hoping that SEGI is a viable solution to this problem (maybe with async updating on a worker thread.) However, if unity managed to come up with a different solution for giving trees better lighting (maybe occlusion volumes?) I'm all ears, too.

 

i completely disagree, realtime GI does make sense in outdoors much more than in interiors where baking can still make some sense for the finesse of it's shadowing.
in addition, baking doesn't work very well in exteriors , it take forever coz baking seems not to be adapted to that kind of scenery.

 

But you get arguably the same contribution from your reflection and light probes, which include usually, the main skybox, so that is all your bounce data for outdoors...

 

Do not see how this would shadow the trees. Would just add the un-shadowed bounce light to the fully exposed looking tree. Reflection probes work fine but Enlighten GI does not.

 

eh!!??, what svogi has to do with terrain system?

 

It brings lot better visuals than plain shadows on terrains outdoors.

 

ah for lighting, i thought we are talking terrain system tools authoring.

 

Ha. Yeah kind of agree. Think we just got sick of arguing about the terrain with no headway so started on lighting. Maybe we should start another thread ''Correct Way to Light Our Dream Terrain' . Do think we should get back on track with the actual terrain in this thread. Thanks rea

 

Just use AO for that purpose... that's not actually GI, that is ambient occlusion. GI is global illumination the light, not the shadow that makes it pop, just the colour.

Also I don't see much terrain there. I see objects, which should probably have GI on if they're static, otherwise they should get the bounce lighting from probes (reflection or light probes).

Anyway it's pointless arguing with people.

 

Terrain light probes array won't upadte with real time lightening.
https://blogs.unity3d.com/2016/10/04/global-illumination-in-recore-our-work-with-microsoft-armature/

 

Alright so the top requested terrain feature is Voxel right? Not exactly a feature I dream of but still seems like the most mentioned thing on this thread. I would want a voxel and terrain hole feature similar to Cryengines:

Seems like it would be better for performance to have a terrain with set voxel areas instead of the entire terrain being voxelized. I played with it years back an made elaborate tunnels that I had to climb and swim thru. Worked great on my old laptop.

So that is part of my number 3 requested feature. The ability to make a terrain hole! One that you can pass thru. Know RTP has this but would like to be able to make a terrain hole included with an Asset Store pack. Do not see this being too difficult.

Then after that main request were:

-Terrain blending
-Better painting of textures, trees, and grass
-Better performance for forest
-Better performance in general
-Road and River Tool
-Also a more exposed system so users can optimize and upgrade to their delight

Let me know if I missed anything and add anything that would be relevant and great to have.

 

voxel objetcs are deprecated in cryengine. too buggy not that useful.
the better way for caves, cliff etc would be to let us use any mesh as terrain object (with design constraints ofc due to UV'ing), in adition with geom blending. those terrain meshes would have exactly the same vegetation painting possibilities.
this way, we could use meshes for river shores IE, caves, overhangs etc... more precise, more interesting visually, more control over tessellation and mesh density where needed.

 

That sounds good to me. I am not a huge fan of voxel terrain systems so this with a terrain hole system would be perfect.

 

Awesome! That is a good to hear. As long as that will work with any mesh type and shader that will be a major help to me.

 

It's possible the new terrain system (really it seems to be more about vegetation if you look at the hint dropped) was late due to dependence with the integration of gpu instancing in the engine, we have heard them more since instancing have become close to release, or maybe I'm wrong ...

 

They're focusing on vegetation as part of the instancing upgrade. Terrain rendering itself is another story, and nothing to do with instancing. The instancing stuff is for speed tree, grass and other shenanigans...

As it's been said before - vegetation is probably the largest part of what makes a terrain a terrain. The existing Unity terrain can easily be used to create terrains far beyond what the average AAA title has - it's just difficult to work with and slow to an extent.

But people don't see it because they don't understand that you need to have objects blending into terrain and tonnes of vegetation... You don't really get to see just the pure terrain in a modern game, usually it's covered with all the things you can imagine

 

Terrain vegetation will not cover the entire terrain, if you target AAA you need a good floor terrain look and shaders.


http://cdn.wccftech.com/wp-content/uploads/2015/03/The-Witcher-3-Wild-Hunt.png

 

...which we have in excess of AAA solutions with RTP. RTP is better than most. You're probably confusing actual terrain with meshes in a lot of these games.

 

I know, the question is : will new terrain system improve on these areas or will it only covers some aspects like vegetation letting other areas covered by plugins ? I am curious only as i am not making such games.

Wrong in many cases

 

You are both right. While some environments have lots of cover most do not. Having dense foliage, detail placement on all meshes will help in some environments but there is still a need for good terrain blending. RTP, ats, and Terrain Mesh Blending have this feature but a simpler more universal built in solution would be amazing. Something like DICE's terrain blending would be great.

Definitely think a working more advanced vegetation system, that is not linked to SpeedTree, is the most important thing right now since that is one thing that Unity and the Asset Store does not have. Would love to see this is 5.5. Would be a reason to update to it.

 

No, I am right. What you classify as terrain is a combination of stuff, and what I classify as terrain is a heightmap mesh. In any degree of logical sanity, a heightmap mesh is achievable in Unity with current terrain.

What it really comes down to for actual terrain is the shader only, in terms of visual quality. A "terrain" in games isn't just a heightmap mesh. It's a great deal of things in combination. What you're really looking for is a shader and some really great textures, then your base heightmap mesh is AAA (ignoring performance for the moment).

TLDR - a heightmap mesh in Unity has no visual difference than any other AAA game utilising a heightmap mesh. It is the shader's job in this case (and textures).

I'm only making a big deal about this because a lot of people just see "terrain" and expect it to be ONE single system that handles it all. This is NEVER the case.

Unity have clearly broken up the "terrain" problem into different problems to tackle, like the base mesh, the vegetation, possible mesh blending, the shading and the performance of it all. To start, we get one of those features as it will be working with any mesh: the vegetation. The rest will come when they are happy with the progress.

 

Anyway for curiosity I have start building my own terrain system, don't expect anything performant because I'm a non programmer. But Unity won't implement anything soon that has planet anyway. It will be fun to compare and come from an informed perspective! Now if I can figures out those LOD stitching ...

 

You misunderstood, i was not discussing technical side of things, i was refering when you said

You could make some game like Journey as one terrain only with no vegetation at all and only coins to collect, it is still a terrain.

 

A heightmap is achievable, but the detail resolution is poor to say the least. When you compare UE4 to Unity, not only is it far more performant it actually looks like the terrain you exported out of WM etc. Although, I will mention I'm not a fan of the autoLOD function in UE, it makes distance parts of the terrain look rather odd.

Anywho, it's so simple to create awesome materials in UE that does what RTP does. Like it was a couple of nodes to create a camera based noise masking function to split up grass etc., there's parallax nodes, tesselation options. You can create materials with holes with little effort, you can vertex paint water (puddles), colour maps and NM's are ridiculosly simple to add etc. with only slight changes to the blended weightmaps..

From some of the examples it shows you how you'd add really cool stuff like windy sand detail into the basic terrain shader from desert examples.. Materials aside, the sculpting / brushes / world streaming / ramp and spline tools work well in UE.

Then when you really deep dive and look at the "world creation" system it goes even further, which has automatic point of origin system. It supports HZB dynamic occlusion, dynamic navmesh up to 20KM2 (last time I checked), you just import a big bunch of tiles and set border boundaries, then everything is auto-magically (it's a word) either streamed or async'd and it does it properly.

Then there's the graphical extra's, instead of cascade mapping long distances that does nothing but eat GPU power they found ways around it with lighter ray traced distance shadows that take over at certain distances. Atmospheric fog (light scattering through planetary atmosphere), exponential heightfog (with in-scattering)..

Now I'm sure it could be made better, but I'd be happy with UE4's system in Unity TBH.. It'd more than do for me.!

Edit: Oh and they REALLY need to fix the damn frustrum culling in Unity, multiple renderstep passes for every terrain tile??!

 

i agree with the part UE4 render the distance part of the terrain, it look really nice.

 

In a vast majority of cases, yes, I guess. Although in some settings, you can't really enhance or "hide" the terrain with vegetation: deserts, lifeless planets, post-apocalyptic worlds, or like in the images below, winter wonderland:


Very happy with what you can achieve with shaders like Colormap Ultra, used for the above scenes. I'm all for performance and a tool-set that would speed up the creation process - and proper integration of SlowTree.

 

Wait that your game? i've seen the development thread. really cool game you got there

 

Who are you?
Indie Artist/Developer - Purchased unity 5 Pro, especially after it received a more realistic Graphics engine..

What kind of game are you trying to build or would like to build?
"Active VR" Photo Realistic Worlds

How does terrain fit into that?
Procedural & sculpted Voxel World with Multi Platform compatible Tessellation/Realtime Subdivision & Scanned Displacement Texturing

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

Many of us are seeking a realistic terrain system, I wish OPENSUBDIV to use with displacement was integrated as an alternative to Direct X Tessellation so it could be cross platform.

I'm aiming for realistic VR experiences and would like to be able to use multiple textures to paint with Tessellated Displacement as well as sculpt the terrain in editor optionally in VR as well. Will be using the Virtuix Omni treadmill to get the correct perspective for sculpting.

Essential
*Support Multiple Tessellated Displacement Textures - 6 or more with brightness, Hue and saturation for each.
*Be able to sculpt caves & overhangs and join other terrain areas like Voxeland or other voxel systems -without stretching!
*UV-less terrains - no texture stretching
*Tessellated trees/bark - Tessellated Trees that can be effected by forces/physics
* Each Texture slot to apply it's own physics and sound and potentially a particle system that can be triggered based on impact strength from jumping, falling or sliding etc -to add splashes, dust or ice spray. This would allow easy advanced creation and setup.
*Quality River system Nvidia Flex -realistic fluids that follow the sculpted rivers and physically effect other objects with collisions.
*Terrain Culling/streaming system built in if possible
*Tree collisions follow area only of player

I Will be using object/terrain streaming tech as well as texture streaming
*Tessellation Displacement shader that allows blending objects on the terrain.
*PTex Support and texture converter if it makes sense for performance and Visual - this also bypasses the need for UVs.

 

that one is actually impossible ...little benefits for huge perfs drop.
i already asked for that several times and agree that rivers/freshwater flows really suck in unity and other engines.
an approaching system for flows should be investigated, maybe bake some river meshes in editor at build-time, with a system similar to light baking, but for flows....

 

PA Particle Field on the GPU is pretty good with performance with up to said 16250 particles with just 1 drawcall .. if a mesh water animation was linked together in optimized sections (small cloned sections), followed a spline system, was streamed and or had distance based culling (macro culling) might be good enough or maybe some kind of micro culling -particles that the camera can see at close range aren't rendered. The macro and micro culling for Flex might also be helpful.

Also about using Flex, I'm using it now for VR with World Streamer and with other optimizing techniques may make it doable. It's still in development but I'm hoping it will evolve to allow better performance.

So then make it easier? and add a easy cave sculpting system like Voxel tech. Maybe the terrain can have surface auto detection when objects are placed so that a prompt for objects comes up with a dialogue asking the developer/artist if they want to add or switch to a shader that allows that object or objects to blend in with the terrain.

 

Who are you?

Damjan Mozetič, an indie developer.

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

I am working on an old-school RPG, Call of Saregnar. See www.facebook.com/CallOfSaregnar for more info.

How does terrain fit into that?

Pretty much every RPG needs a terrain system and mine is no exception. I've been struggling with the current terrain system as I couldn't create a polygonal look in any way. My solution was to model many square chunks of terrain (plains, roads, hills, mountain pieces) in Blender and tile the models together to create a terrain.

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

I would like to be able to have some flexibility for creating and sculpting old-school-looking polygonal terrain, nothing more.

 

Being on the team that develop Landscape Builder, we think there are many great tools in the store (obviously I'm biased towards LB). Having a healthy tools community is what strengthens Unity as a development platform. We all have different approaches to solving game development issues. Out-of-the-box tools are much less important than core, low-level functionality. Performance of these low level terrain features is paramount. Providing open, flexible, performant APIs in the terrain system will:

• Enable tool developers (like us and others), to continue delivering tools to speed up your development and provide great looking landscapes
• Enable anyone to build their own solutions on top of the core Unity terrain system

So, maybe the question is really what core features do we need? It always seems to come back to wanting more performance. Which is why vegetation and texturing are so important. With more performance can come more features which can be used to improve the overall look and feel of the terrains.

Like with our image fx shaders we can sometimes be asking ourselves, how can we get back that 3ms so that we could do XYZ? So it is with the terrain system, if this core component was faster, then maybe we could do .... (fill in your favorite terrain feature here).

 

a decent terrain system IS a core feature.
we are not all hardcore coders, otherwise, we'd just develop our own engine.
yes you're biased somehow. you are kind of affraid to loose your utility, wichis understandable.
butthe core functions as you say will remain available so you can still develop your own tools and assets.

 

Who are you? What kind of game are you trying to build or would like to build?
I'm Alex, one of the creator of Kona. We are trying to think of how we'll make our next game.

How does terrain fit into that?
We want our worlds to be big, snowy and awesome; and the terrain system is the easiest/fastest approach to create those.

What use-cases, features, workflows do you have or would like to see?
Right now, we use big terrains that we split into a large number of smaller ones, then convert them to meshes that can be packed in Asset Bundles and streamed around the player. The terrain simplifies the drawing process of our production, but it has a pretty high overhead on the hardware and can't be easily streamed, hence the current need to split and convert terrains to meshes. Therefore, we would love to see a native terrain splitting and streaming solution.

To build our roads, we had to align spline over our terrain and then project those splines to generate our road meshes. This process wasn't an easy one and we currently have no easy way to build our roads faster and in the way that they blend beautifully with the terrains. Therefore, our mind would be blown if we could have access to a native drawing function that works with vectors such as lines, circles and curves to make building roads faster and more efficient.

To build our caves, we use standard meshes. Simply put: It would be nice to have a terrain system that supports caves.

 

Any more official news on progress?

 

Yes, more details since 2014 would be nice!

 

I would like access to the low level vertex generation and for whatever toolset you paint with to be customizable.

My asset, MegaSplat, allows you to easily blend up to 256 unique textures in a single pass with height based splat mapping (with less performance hit than Unity's terrains), but I cannot make it work on Unity Terrains because I don't have access to the actual vertex data (which requires marking and splitting some verts). I have my own painting workflow, which is much more advanced than what Unity provides because when you can have hundreds of textures instead of just 4 or 8, you can use multiple texture to break up tiling rather than resampling with different UVs and other tricks, so my brushes allow you to blend multiple textures together with noise, not to mention access hundreds of textures instead of just a few. So ideally, both the generation of vertex data is customizable and the painting workflows are customizable so we can adapt assets like mine to such a system.

 

Having just started to dabble again in adding grass and detail meshes to my newest prototype, I was positively surprised to find the Unitygrass to suit my use case better than all the way more modern grass solutions on the asset store I tried.

Don't get me wrong, the Unitygrass system is dated, and some of the grass solutions on the asset store are really great, but for my use case, unitygrass did what no other grass system was able to do: render color balanced grass that integrates well with my grass terrain material instead of standing out like a sore thumb, and antialiases pretty well with a good AA algorithm, which often is difficult with more elaborate grass shaders with normal maps, specular highlights and all that stuff that potentially leads to further aliasing, expecially when looked at from a distance.
Lighting works like a charm, didn't expect that. Guess the Unitygrass is taking lighting information from the terrain lightmap, which seems to work fine even in fully dynamic lighting... at least for simple cases like mine.

I often badmouthed the old unity terrain system, but for once I am full of praise. Has to be said once in a while.


Now, I still see a ton of potential to make the grass system, and the detail mesh system more useful for me, and some of these will be essential in the end for my prototype:


1) I think its already on the roadmap, but just to stress it again: Painting grass on all meshes, not just the terrain. Really urgently needed. If its too much of a hassle just keep it to static meshes, though I could see some uses for non-static meshes (monsters with grass growing on their back for example).
In the same tune it would be cool to finally get one system that would allow to paint splat maps on ALL meshes, not just terrain, and synchronize the materials used between all the meshes.

2) More specific, having grass that can vary in size between single grass instances, not just between patches of grass. Currently each patch gets a height assigned, and each billboard of the same type in the patch shares the same height. Would look much more varied if the height would vary more between single instances.

3) Better wind movement. Have a look at other grass systems, that actually implement "grass wave movement", and not just propagate the exact same vertex movement to all grass instances all over the map. Having the gras move in waves would make it look way cooler.

4) shaders. Really, give Unity devs access to the grass shaders, so they can write their own shader easely and replace the default one without having to find the hidden shaders.
Also, a selection of default ones for people that do want specular highlights, or normal maps for their grass for example.
Same really about the detail meshes. Just a vertex lit diffuse shader does not the trick anymore in 2016 in most cases, especially when everything else is now using PBR.

5) More than just Billboards. I found that billboards work best for my use case, but I guess many others want more complex grass. Options are always welcome. Best case you could mix the different types per grass type.

6) Shadows. Now, I am not sure how much use shadows are for grass. I often see bad examples where grass is shadowing way to much for the light transmissive things some strands of grass are. But again, the option to use realtime shadows for the grass and finetune the intensitivity would always be welcome.
For detail meshes, this option is essential. Really. Small detail meshes without shadow casting look so weird next to normal static meshes casting shadows. Really immersion breaking.

7) Better control for density of placed detail meshes and grass. Currently the "Grid" on which these are placed seems to be uniform. It would be cool to have the ability to get a much higher density at some places without having to increase the global resolution. How that system should work IDK. Still, something that would be useful.


Well I will not go into trees or the actual terrain shaders / materials, or terrain system. Trees are currently WAY to broken to even know where to start, and I am pretty happy with the terrain shading / materials that thirdparty solutions from the asset store provide so I have no real need for a better system from Unity there, though I will not say no to it

 

I want to add a subtle one:
- Remove the dark gradient at the bottom of grass, it should be optional
I wanted to make a grand cartoon scene with solid colored grass made a white texture and using unity's default coloration .... didn't worked because you can't have grass without the gradient so it killed the effect completely ... I mean seriously, the artist is responsible for that not unity, unity should take decision for us, especially minor visual one like that

 

In the meantime, while we wait for the final Unity terrain improvements, I, finally, managed to fix all the issues with my SpeedTree system I've been talking about. It is release as 'final' (maybe even production ready) for Unity 5.5+.

It goes absolutely great with tile-able terrains, hundreds of thousands of trees, multiple tree types, smooth lod transitions, instancing, lod transitions working in instancing too, CullingGroup API, 300M+ view distance for mesh trees, with infinite billboard view distance, and all free + open source!

Using an exterior tool for generating your trees is also possible (like Gaia) since, of course, this is just a little system built on top of the unity Terrains..

Check out the forums and the demo/repo if interested!

Our results with this system for our upcoming open-world game have been far beyond our expectations for Unity:

Quite dense, right?

 

@superpig Please allow for "cascading" grass instead of a single cutoff line. There should always be grass in the distance, just less of it.

 

It would be terrific if we could get some official update on the status of the new terrain tools, given they are still in the roadmap under research, thus I guess this means "computer says no" for a 2017 release... does it?

 

New Year, New News?

 

This is not grass in distance , but shrubs that are less dense and that displays farther, it is like a second grass system.

 

happy new year 2018 maybe ? grin

 

There is plugins if you need advanvced terrain features, why waiting ?