My experience and thoughts on HDRP since its release.

OK, big thumbs up here for the roadmap talk taking a moment to show which features are planned in LWRP, HDRP and Shadergraph during specific Unity version cycles. This was one of the things I asked for when moaning about some aspects of the GDC Roadmap talk.

Also thumbs up for mentioning the new VFX system, its now scheduled for preview in 2018.3.

Likewise terrain did not go down the memory hole. 2018.3 is also the target for a whole range of terrain system & tools changes. Quite a bit more detail about performance side of things compared to the last roadmap talk, though it went by so quickly I havent had time to absorb all the detail yet.

So, all good from my point of view, and much of my fretting was needless after all, yay.

 

Of course your fretting was needless. Except for the bit about forum formatting...

 

Well I am bound to say that some of the fretting is an inevitable side-effect of the roadmap on the website not always being kept up-to-date and complete. So I shall be watching to see when all the stuff clarified in todays roadmap talk makes it to the roadmap on the website

 

Missed it live so I'm watching it now always excited to see what the roadmap talk holds, gotta take some fretting from elbows waiting for the new stuff

 

I wasnt fretting that much until someone else started paying attention to the roadmap RSS and went on about it here, honest!

Anyway, this is the HDRP-relevant slide from the roadmap talk:



There was also terrain and VFX stuff that is relevant but I dont want to go bonkers with screencaptures in this thread. I did post all the terrain-related ones here: #15

 

I've been eying the mesh decals on GitHub for a while now so I was glad when I saw a timeframe for that

 

Does anyone know what Stack Lit Shader is?

Also, I wish GPU lightmapper was on that list...

 

I think they tend to talk about that as its own feature, not as part of HDRP-specific roadmap.

As for Stack Lit, I cant remember what info (if any) is already out there about this. Will have a look around when I have some time and see if I can spot anything. Think it was mentioned to be a cinematic quality shader in the roadmap talk, but I am paraphrasing since I dont recall exact wording.

 

Thanks for the reply. I just have no idea what "stack lit shader" is. It is probably something we all already know...but just can't make the connection. Or maybe its a brand new magic shader

 

It's the cinematic quality shader as elbows mentioned above and its also what's stated in the Roadmap talk at Unite Berlin

Only other info on it so far

 

It's for cutscenes and movie generation or totally obnoxious rigs It's pretty much useless without a whole bunch of AAA art thrown at it ...

 

During Unity Berlin, when they were talking about unity's roadmap, they literately ignored and was so whatever about the gpu lightmapper they been teasing forever. Like no info during the the coverage of it's roadmap . Very disappointing.

 

Well there's more to come but the last I heard it's a 2018.3 or later thing.

 

I am not sure why the information about the GPU mapper is a closed box. They demonstrated that it works to a degree with that video, but just haven't released further information about it. I find it very strange. Is it being tested among closed testers at the moment? or are we far from that?

Some indication of progress would be really appreciated

 

I'm kinda worried that no infos about surface-like shaders and instanced indirect support has been provided, is it even worked upon ? Tons of assets uses it and migration is impossible without that.

 

Because there's over a decade of examples from Unity users trying to stir up trouble and "hold unity to ransom" over deadlines and design.

For example well-meaning staff will talk about things they plan to do but that's just it - it's a plan that might shift with a deadline. It might be changed. Stuff may be added. Removed. Each time that DID happen in the past, I, or a community manager or another mod had to wade in and sort it out because people don't understand the difference between intention and implementation.

What is more, in Unity land, the users often don't believe there's this concept that you can change your mind.

Most people here read what they want to. They'll read "terrain improvements" and take some vague suggestion from 5 years ago and wail wildly that it's not in 2018.2! it's not in 2018.3! and so on. That really does happen.

This is why - no - Unity shouldn't be forthcoming until it's done and almost ready to go beta. Simple as that. Believe me it's been tried many times to find a balance and it doesn't exist because people aren't reasonable and attack Unity if they change their minds. So the roadmap was born - and even that has to be carefully changed.

Here is another reason it doesn't make sense to go public too soon (although staff certainly do not ridicule users, it's just for fun):

Spoiler: Joke

 

Having said that, whoever is leading a particular area can, of their own accord, choose to add information they feel is right to add - it's entirely up to the staff, I'm just adding a bit of background *why* sometimes they might not!

These days, Unity is a really transparent beast. In particular with the ECS part, they're very vocal that it's "WIP" and will definitely change over time. That's fine too!

 

I have a little bit of sympathy about GPU lightmapper not being mentioned, because thats what I was fretting might happen with other features that I care about that have obviously had their timescale slip a little. And you know that in my case I dont really mind delays etc, I just mind lack of communication.

In the case of GPU lightmapper I am entirely unsurprised if the schedule for it slipped. They really havent teased it forever, and towards the end of last year it was described as being in a pre-alpha state. In a Unity blog post at the end of March, they said it was expected to be released later this year. Granted, in the GDC roadmap talk it was another of the big features that they attached to a version number (2018.3) that didnt stand the test of time for very long at all. Given that terrain went from 2018.2 in that talk to 2018.3 now, and that VFX system went from summer preview to 2018.3, it wouldnt surprise me if lightmapper had slipped to 2019, or if they just want to be cautious about timescales with it (eg they might well have a small, closed test for select people to try an early version and only with that feedback get a sense of how ready it is to share with a broader audience). I also seem to recall that some of the language used when giving a 'sneak peek' of it at a conference tended towards the idea that it might first appear in a special separate preview build of unity, again in order to receive more feedback before attempting to put it into a mainline unity branch. They hyped up its capabilities, but not how ready it is IMHO.

On the tech side of things, I suppose I do have to wonder whether Apples decision to deprecate OpenGL for macOS might throw a spanner in the works, given that the GPU lightmapper seemed to be OpenCL-based. But this is just speculation and since I havent touched OpenCL for years I'm not really up to speed on quite how this relates to OpenGL.

 

I agree with you. I just want GPU lightmapperso bad

Apple users can stay with the CPU bake...and enjoy it

 

I think that time might have arrived. I'm so happy with HDRP, some other planned features and Unitys communication of late that I dont think there is much room left in my life for any other engine. Well, really I'll probably still keep half an eye on some other things but Unity is likely to go from 95% of my engine time to 99.9%

 

I seriously considered working on either CryEngine or Unreal for my next project. I almost did to be honest, but as soon as I played around with the HDRP, I immediately realized that Unity now and 2 years later will be very different, so I stayed

 

It's always been mint if you know how to prod it.

 

Hi there. I want to share some experience of using the Layered Lit shader. I wanted to blend several layers of wall texture with vertex color blending and I faced with situation in wich I cant clamp the blend result between one absolutely flat heightmap and another denser one. Result loosk like this:

I tried the extremely high height numbers and the best result looks so:

This is very strange technique. Just because I must agree with that my texture height is 10 meters ))) Maybe this can help someone.

 

I believe it's much better to look at what games are actually released with Unity and how they look, rather than Unity's PR showcases (and then be super greatful when developers actually release their solutions, like Playdead's TAA, for example).

Unity's Blacksmith demo, with screen space "character shadows" was three years ago, and only now in HDRP do we get an official screen space shadow solution. Unity's first Adam demo with shadow casting real-time area lights were two years ago, and only now in HDRP is there apparently an official area light solution (though I haven't seen anyone test this yet).

In other words, Book of the Dead probably contains as many "hacked together" custom solutions that only works in their highly specialized use case, as all of the rest of their demos, and you shouldn't expect to be able to use any of it for at least a couple of years.

(Meanwhile, Unreal has looked like Book of the Dead for ages, so I'm curious as to what else actually drove you to Unity... but perhaps that's a different topic.)

Personally, I wonder why there aren't more realtime GI solutions yet, since apparently all AAA developers with custom engines have switched to that years ago. But then again, there's none available for even Unreal (unless you count Nvidia's solution that's not really performant, and restricted to their hardware).

 

Realtime GI solution was in the CryEngine 3 even in 2011. It works like voxels, resolution was 1 squere meter, so it was hard to use in things like architectural visualisation because voxels are bigger than your wall depth. I don't realy understand what's used in Cryengine now, but it's realtime GI looking like raytrace, maybe it's voxels but screenspace instead of worldspace.

Enlighted is available on Unreal, but it's not free to get just because Unreal is opensource, but Enlighten is not, and there's some copyright problems I don't really understand.

As for Volumetric Fog there is one indie finland developers Theory Interactive created their inhouse engine Practice Engine for the game called Reset.

I'll be happy to use something like these ToD simulation. It's volimetrics is totally amazzing!

 

I'll list the main customizations to the HDRP for the Book of the Dead Environment that got released so you understand whats really custom. Keep in mind HDRP isn't finished yet so things are still being worked on to be implemented.

Major

Sky Occlusion Probes - Maybe coming to HDRP at some Point ^

Atmospheric Scattering - Coming to the main HDRP (Whats in the Demo can be achieved with Volumetric Fog)

Grass Occlusion

Custom Wind System - Main HDRP has support for wind. I remember seeing somewhere that it will be improved.

Minor (Specific to the project/some things anyone might/might not change for their project)
Shader Terrain Support - It's using a custom Layered Lit Shader with support for terrain rendering (Main HDRP support for Terrain is coming)

Custom Lit shader - to support the Wind system and Trees with a second detail texture/parameters (For the treetops - You can make your own version of this with Shader graph)

Area Volumes - They just added lights, Atmospheric Scattering, and a few other things to the volume system (Scene Settings) So they can blend their properties for different areas. (That's why you see the light/shadows move in the videos)

There are some other smaller changes/custom things but that's pretty much it. The main feature everyone wants from Book of the Dead in the main HDRP is Sky Occlusion Probes.
So currently you can do everything you see except having that detailed wind system, Occlusion probes, and terrain (officially since there's a hacky way to do it). By the time HDRP is out of preview/released, the only thing missing might be the Occlusion Probes.

 

Sounds like a bug.

 

Hey there, I have a success! I made my own vertex lit shader to create the damaged wall surface


My solution allows me to use one vertex color channel, in fact the red one, to blend between several stages of surface damage. I use the same height mask i used to make the cracked paint in SD and make some histogram scanning to mask all the stages and blend them together.

My pipeline was to make the prototype of a shader in Substance Designer and then reproduce it in Shader Graph with HDRP. It happened to be not so similar and not so easy, but result looks just what I wanted to get.

 

Just curious, wouldn't decals be more practical for adding that sort of detail? Or are you just testing it on the wall?

Either way, good progress!

Occlusion probes are a must! Anything related to lighting is a welcomed feature honestly.

 

That's not the sort of effect to be done using decals. Vertex Painting saving your batches and gives you variety if details if you using techniqhes like height blend. I'll use the decals to make the leaks, labels, maybe some damage with parallax to the walls. Will be cool to have the tool like Road Tool to make some long marks on the floor

I'm not so happy with Occlusion Probes. I love the scenes with ability to change the indoor/outdoor with day|night cycle. But occlusion probes uses the progressive lightmapper, it gives me tonn of lightmap artifacts in the indoor scenes with more baking time against Realtime GI with Enlighten with no artifacts at all. I'll be happy when GPU lightmapper be released. So in my case the Occlusion Probes are useless )))

I'll be happy with some VxGI solution out off the box

 

Didn't realize it was for vertex painting btw, how would the vertex data be stored? Are you generating texture for each wall with different vertex paint data?

nVidia announced VXGI 2.0 and I wish Unity embraced it.

 

VXGI 2 features require Nvidia Maxwell+ GPU, meaning it's not going to be a viable option for a commercial game since you need to support other/older GPUs as well.

 

If it happens, then it will be ported to a compute shader version for reasons you mentioned. Also, I think it is already out of the project scope if a game with VXGI or any realtime GI is fully supported on older GPUs. DX12 and SM5 will be required anyway.

 

It's not the way it works. I have just one material for all the same walls. Vertex data is a set of float arrays: position (x,y,z), texture coordinates (u,v), skin weights, vertex color (r,g,b,a), and other I don't know yet. If the mash has vertex color set, each vertex has it's own color. You can grab the vertex color data in your shader and use the way you want. For example it's the simpliest way to paint your lowpoly stylized model

Or you can use vertex color to blend between materials in your shader. It works like terrain does.

You can use the ready tools on assetstore like VertPaint
https://assetstore.unity.com/packages/tools/painting/vertpaint-100282
Vertex Tools Pro
https://assetstore.unity.com/packages/tools/painting/vertex-tools-pro-55649
or use the vertex paint tool in the ProBuilder package now available for free.

 

Cool. Unity3d's HDRP LayerLit use vertex color to blend different layered materials too.

 

I see how it works now, so I assume, to have same mesh with different vertex paintings, you would have to have different materials or a separate data holder? Or am I still missing the whole thing XD

I ask because I am not really familiar with it.

I am thinking if it can be applied for small terrain (maybe 3 x 3) details. Just curious.

 

If it's mesh terrain then you can use it for that and you won't need to make a separate shader in Shader graph unless you want some other feature the Lit shader doesn't have. Otherwise, the Layered Lit shader will work perfectly for this.

 

That is what I am doing now but I am just looking for alternative solutions.

 

Well... Layered Lit doesn't work perfect for me so I wrote my own shader in Shader Graph for that.

You can have several copies of one mesh painted defferent ways right in Unity using some Vertex Paint tools for that. And you can apply one material and have different results with different vertex paints.

 

If you want a different vert paint shader then you'd have to use shader graph as xrooshka did. Otherwise, actual terrain support isn't too far off now.

 

Create custom HDRP shader using Shader graph ? I wonder shader graph can well support custom HDRP shader now. do u using shader graph version in Github?
thanks.

 

Shader Graph for HDRP is available from the Package manager just get the latest version. Also, this is 2018.2 forward so if you don't have 2018.2 yet you'll have to get that it got released today.

 

2018.2 and latest HDRP updated to 2.0.4. I take out baked global illumination
I've got two suns. One is the sunlight component and I think the second one is backed in the reflection.
Any idea? looks like the reflection is in a different direction

 

Keep the current terrain system? Are you having a laugh?

 

Just because it is outdated, doesn't mean it has to be deprecated immediately. Until we a new version fully ready, having support for the old version will help immensely.

 

Terrain/outdoors quality is the major component holding Unity back for my projects.

Planning out my product roadmap, I'll be going for a simpler 2.5D (in 3D) strategy game to be launched with 2018.4 in year 2019, and hold back a grander strategy game with realistic outdoor visuals until 2019.4/year 2020. From reading the roadmap and this discussion, it looks like the full suite of tools and new features won't be ready until the 2019 Unity releases.

 

I suspect the window of opportunity where such an intermediate step may have been particularly useful and timely has already passed. Because although it is still too early to tell for sure, it seems like initial efforts to make HDRP work with the new terrain system are well underway (ie a 'GPU terrain' branch on github that is at the review stage and has recently been updated to take account of the first code review). Obviously until we get our hands on a 2018.3 beta that actually has some of the new terrain stuff in it, we dont know the detail. But even if either the new terrain side of things or the HDRP support for it are initially incomplete in some ways, HDRP as a whole is not considered to be close to fully ready. So since terms like fully ready dont yet apply to the system as a whole anyway, I wouldnt be looking for fully ready old terrain support anyway, it would just have been a slightly odd cludge.

I know there are some scenarios where support for the old terrain system would have been useful to some people, but personally I'm glad Unity have taken the forward-looking approach to this stuff and not used their resources to attempt to support the old ways. Especially given how old and 'flawed by modern standards' the unity terrain system is, and how long the process of developing HDRP is as a whole. Again this is easy for me to say because I have the luxury of following along for the ride without having any deadlines for my own stuff, but I'm really enjoying this era and although I'm a little nervous about the new terrain right now, this is only because the public reveal of this has been quite limited so far. It was certainly great to see comments from Unity in the blog post about the roadmap that indicated that the Unity terrain team has expanded quite a bit this year, and I am pretty excited about 2018.3 (and whatever may actually follow in a later release).

 

You need to make sure you use two different volume profiles for hdrp, one which includes the sun, and one without.

 

I

I'm using the default scene. To be clear, I'm using only one default directional sunlight. The sunlight component is one white dot and the second dot, that is in the shadow, came from the reflection. I will experiment more with different volume profiles for HDRP.

 

Correct me if I'm wrong but I think the design is:



You have volume settings but the baking profile should have the sun rendering removed, so it doesn't show in the baked skybox.