Ocean Renderer

Hmmm, mind if we build on it? I've not had a look at the project files, I'm sure by reviewing / modifying / adding vertex displacement, Fresnel bias and FFT map it will be horrible performance wise for mobile. But it looks like it could be very handy for a CryEngine style water..

 

Click his blog and you can see it all in action and how it's done

 

I want permission to improve it an use it LG..

 

When I say dx9 what I really mean is openGL3. I just got use to say dx9 to mean 'not dx11'. Confusing.

So yes it should work on a mac but I have not tested it. Let me know if it works please.

I havnt had a real close look at the code but I think the clouds are done with a cheap method that's works quite fast. I know that there is a very expensive cloud render done by Eric but I don't think it is used.

Yes the tree's are going to be hard.

Yes, its is open source released on the GPL licence. I am always glad to see people extend it and from what I got from Eric and Fabrice is they feel the same.

In terms of the Proland port there are issues with using it for a commercial product but it can all be resolved. I will cover that when I finish the port.

not sure what LG stands for?

 

I believe it works fine for MAC, but not so well for Linux,

 

Scrawk, i think he meant LG like lazygunn, and yeah, i didnt really want to say it myself but yeah ShadowK i've already used and adapted scrawks whitecap version of his brdf ocean shader myself, you've seen a video of it in fact

Oh, additionally, while I now seem to have a lot on, i'm going to continue to figure out the whitecap shader for my uses, i think you said it doesnt really matter what grid it uses so i'm going to try a square grid with a mind for it more easily going into the suimono beta i added sections of to, then it would sooner go with suimono's general system with tesselation and the suimono generated flowmaps for coasts

Probably going to take a bit of figuring out but i'll see, at least i got closer enough to understanding it to add generation (takes a moment so cant really be done realtime) of spectrum textures and blending between them - this has no effect on performance it seems so as long as you dont need loads of states its fine for as you first described, transitions from calm to stormy and back

 

OMG, wait a minute. Do you plan to redo Proland with Unity. Proland?! with Earth?

 

Sounds great, we have an ocean shader / toolset for an engine we built (Well partially, it's buggy as hell) around the Crysis 2 era, it's just a matter of converting the code and adding to what's already in there.

I did mean LazyGunn ..

Performance wise it was never aimed at mobile, but for high end PC and Mac's.. So it'll be interesting, thanks .

 

Yes. Im going to give it a try. Should take around 3-6 months.

 

I've used this for a few Simulations and client demo's in the offshore oil field.

Works quite well and consistently.

 

Nice. Very nice. Good luck et bon courage.

 

Depending on which version you use can make a difference with a few things - You don't need to add your own fft generation, its used a bunch, and wave shapes are very convincing at a huge range of scales. The whitecap version, which is the one I use, is a bit of a beast, with i think it was 9 fft operations? Performance is not an issue if you can run it on the gpu, however, even my whitecap version with a bunch of extra shading processes going on that the suimono creator was helpful with runs at 300fps on a 580. The brdf/whitecap oceans use JONSWAP spectrum but he's done an ocean using a Phillips spectrum implementation which is a bit more like the wave shapes you see in the original crysis.

Today i'm going to look at the dx9 brdf system because i'm very interested in its performance but my main goal is to get the oceans working with a square grid so they can work more intuitively with suimono's system. I think i've subjected scrawk to way much hassle so i'll quit bugging him and it'd do me some good to work that out, and i think i was overcomplicating everything because I was stuck so much on trying to make everything work around a radial grid

Converting Proland is so amazing, i'm really looking forwards to the results of that. The trees and ability to scale detail like that would open doors for many projects

Anyone who enjoyed the videogame Frontier many years ago could probably recognise the significance of Proland working well

 

Just as an addition to your flowmap implementation on your blog, might something like this http://www.polycount.com/forum/showthread.php?t=105399 be useful for painting the flowmap? Probably not as accurate as the unity offereng using fluid dynamics to create the map, but that costs money

And yeah, if yu get proland's tress figured out, ill be head first in your code like a man posessed, but ill spare you from unreaonable support requets

 

That looks like a good tool for flow maps.

I have just been setting up Proland on my PC the last few days. The 3D tree look a bit strange up close so I dont know how useful they will be. There are also 2D trees that look fine for billboards.


They physics effects project that I have been working isn't really working out how I planned so I think I might leave that for now and instead do the underwater effects project. Just started today but already looks quite nice. Here a screen shot. Still lots of stuff to add so final will probably look different.

The suns setting so its a bit dark.

 

That's looking top notch, well done, by physics do you mean bouyancy and suchlike? I'm giving up hope for that with whitecaps and just not putting anything floaty on it. Getting the waves to attentuate/adapt near a coastline is another thing entirely, hoping using a tesselated rectangular grid will help in my understanding in all that (as in move the grid vertices first, then tesselate), if only for it to fit more nicely into suimono's system. Will keep following your work! It's always very interesting

 

Wow.....that's looking great scrawk.
Oh btw can anybody showed me how did you create your texture lookup? something like luckymouse made for sky lookup texture.

 

If you mean how the transmittance, irradiance and inscatter textures where created then they already come with original code from Erics project. No code was provided how these where made but the process was outlined in his atmospheric scattering article. The inscatter is quite complex to make and Im not sure exactly how it was made.

EDIT - looks like there is source code for how these are made. Its in his atmospheric scattering project source code. Don't know how I missed that. Im not sure if that's what you where asking about though.

 

let me know if I can help. the original code was written using code blocks, but I use visual studio 2012 so I first ported it to that and I have the solution files avail if you want em. I have it compiling etc

second, the code was never set up to be production quality so there are a number of issues, errors and architectural 'surprises' such as certain 'liberties' taken with pointers including their own smart pointer implementation which has been giving me headaches for weeks now..

it was basically set up to test sets of features to prove out different math treatises, at least that's what i'm getting out of looking at the code for the last month.

besides the fact that ork is their own renderer, there are a number of things to consider: using their own renderer is more of a 'push' system vs. unity, which you might consider would be a 'pull' system since UT calls back (update(), lateupdate() etc) which creates issues for threading/batching..

if this is to be a 'true' port then it won't be as simple as just compiling a dll in c++ and using it that way, since using external dll's have a lot of restrictions. it'd need to be a true c# (for me at least) port. that means:

deconstructing ork, their templated vector, matrices, box etc structs (I have this part already done, although I haven't tested performance yet)

scheduler
scenegraph
resources
tasking
threading
lang conversions (templates, multiple inheritance etc) c++ to (c#, javascript, boo?)
glut/glew stuff
shaders. this will be a lot of work all on it's own since it's GLSL.. a lot of the pre-compilation is cpu bound

basically you can spend a month (or more) just trying to identify where you can separate functionality and much more time on straight conversions and the issues that come up with that..

then there's the big one I hit.. precision.. proland has it, unity doesn't.. converting everything to floats makes it nasty.. there is also a lot of pointer math and other things that don't translate well into c# (or for that matter, javascript)

to be honest, I don't really think it's going to be possible to do a straight port. and even if you were able to, with the loss in precision you'd have a lot of potential artifacts.. some visible, especially during patch culling/creation etc..

I've come to the conclusion that you'd need to basically write something from scratch and just use proland for reference only, than think of a port..

just my advice.. if you need anything, just let me know and I will share what I can

 

The previous post definitely left me feeling a bit stupid and now this question is going to seem a bit narrow or dim, but basically, can you reproduce particular proland features individually or does it all rely on being a whole? The reason i ask is - trees. I finally found a good fit for a project in terms of location that is 200km x 400km basically taking a big chunk just east of seattle, and i'm going to use world composer for grabbing the elevation and satelite imagery for building the terrain. First thing I noticed when looking at the satelite imagery was that there were a hell of a lot of trees. I'm not well aquainted with america but I didnt expect there to be so many trees, but also remembered all those trees in proland and my brain lit up a bit

It would obviously force a certain aesthetic using 'prolandish' trees, but things like that and even the way the ground is shaded seem like useful things to take from it, and there's no way i can do it i'm an imbecile when faced with such things, but i sure do wonder if proland can offer unity users an answer to their 'how do they fit so many trees onto a mountain' dilemmas terrain folk might face - given what it may imply aesthetically

I suppose i'm just asking someone to do something i cant do for little reward other than an internet thanks, but if you like thankses id love insight into those trees

 

Hi Marionette, thanks for your input.

Yes, I have seen some of the issues. Not so much in the code but the whole program is not really complete. Thats fine with me, theres plenty of good stuff in there as is.

It will be a true port. All the code with be moved to C# and the rendering will be done using Unity's pipeline. I have already ported over the ocean and atmosphere code. The shaders have been moved to cg from glsl. I have much more experience with glsl than cg.

Any thing that can not be ported over will have to be rewritten in a Unity friendly way.

Yes, I know. Thats going to be a issue that I dont really have a answer for at the moment. I wont know how bad it will be until every things running. Worse case is that you will have to restrict the world size. Theres still lots of great stuff in there that doesnt need to have a massive world size.

That may well be the case.

 

It doesnt need to be the whole thing. Proland is broken down into separate parts for the trees, ocean, sky, etc. If I can not do the whole lot for some reason I will just do what I can.

 

That sounds pretty cool, im trying to avoid overbadgering people to do things so i'll be looking forwards to seeing what you get up to sans badgering! In the new year i may ask reserved unbadgery questions about atmospheric scattering when I have the suitable terrain in situ and the trees, well they'd change the entire direction of my art but its pretty much something i'd welcome over satelite imagery, and things do get a bit boring going for realism non stop, and to have that many trees, ahh, it would be pretty fancy. There are probably a bunch of complications regarding making them shadow casters or giving them collision or aligning them with a unity terrain surface (all guesses) but yeah. I'll pay you in internet highfives or something.

I think i fell in love with your whitecaps ocean, in the right context it looks unbelievable, so now i refuse to let it go (and it has to feature somewhere or other in the scene for sheer frivolity) however difficult it might be to implement things, or just waiting to see what you come up with next and crowbarring it all together, keep up the great work!

 

I suppose a shader down approach might be best which is how I was going to attack it next instead of from the resource loading up.. it *should* reduce most, if not all of the visual components, to separate modules, however there are a lot of things that won't translate over to c#. Unions for example.

I also ran into a lot of headaches trying to get around things like tuples or specialized lists. .net 4.0 has them, but the version that unity/mono uses does not, which means you'll have to write your own implementation of things already available in the current .net framework.

there are also things to consider like 2x2/3x3 matrices that the proland math classes use that unity has no concept of.. afaik unity only has a 4x4 matrix so you'll have to manually convert.

all of the water/tree/atmosphere shaders are cool n all, but I was looking forward to getting ahold of the actual planetary terrain functions, threading and rendering.

i'll definitely keep an eye on your progress as I have shelved my implementation for now..

 

btw, lazygunn, you have to remember proland isn't just 'proland'.. proland is something that was built on top of their renderer 'ork'.

so while the shaders might be able to be converted, using them in the same way proland uses them, might not be such a simple task..

 

Ahh yep, ive watched a bunch of the videos, particularly the trees (crazy eh) and it's not straightforward, its hard to say things to people who port or create these things for the pure achivement or enjoyment of it that 'i really like those trees how do i get that free', i guess its just probing for anything that will work to someone not wholly ignorant, but certainly not able to hold a sustained informed conversation on the mechanics of it. If its just not going to work with anything else unity integrates i guess im just curious as to the fact.

I was a bit bouyed by the ports of the ocean system because i think the whitecaps version wasnt really shown at its best outside an 'arty' environment, when inside it its a spectacle, so yes, just curiosity based on work done so far, especially putting things in a games context

I think its kind of whats been expressed, theres a ton of interesting things in proland, as chiefly an artist and programmer second my priorties are the inverse of someone interested in the meat of proland, i just like the millions of trees or anything pretty, i would like to knowif thats doable outside the proland context, what i would have to learn etc, because its stuff i can put in a game and it be very enthusing to the uninitiated (i presented the whitecaps ocean to a bunch of friends and interested gaming and art folk, giving necessary credit (lke i just mashed two peoples work together) and people have just fallen over themselves saying how beautiful it is, so its hard not to see potential in this stuff in a further context)

So yes if you think i'm being oblivious let me know, everythings a great education, especially when after many years of being 'just an artist' ive tried to learn more about the technical implementation (And i'm aware that both sides can find themselves hopelessly lost at sea at times when non neophytes take such knowledge for granted), but i do encourage this stuff, although i try to keep it to a forum post rather than direct contact hassle, because i rely on good will, as i have no money

 

I don't think you're being oblivious at all

I just wanted to make sure that you and/or others that read this, understand that just implementing shaders for a particular demo or functionality is often oversimplified and the inner workings to make them work properly are often overlooked.. for example:

if the ocean demo looked great but ran at 5 fps, would anyone really want it? to make sure that the shaders are fed properly requires a certain framework with the horsepower and precision to feed them, ie: threading, memory storage/usage/caching, proper resource handling etc all of the things *behind* the shaders that are cpu bound.

before proland (the api that is layered on top of ork) was even able to process those shaders, their rendering engine had to be able to provide the functionality to do so. thus they spent a good amount of time writing their own renderer. the problem with a port is that it in no way resembles how unity works at the scripting level, proland for the most part is pretty tightly bound to the renderer even though they use 'plugin' dlls..

with a port you're talking about converting struct types, classes semantics not just from one language to another and all that that entails, but different graphics api's as well as shader api's. then there is the renderers themselves.. there are some things that unity just can't do that ork/proland can.. precision is an example..

ork was specifically written with these types of functionality in mind from the start. it's a pretty specialized renderer (spherical terrain), whereas UT is more generic.. there's no guarantee that even when everything is pulled apart that unity will be able to render it properly and efficiently enough to make it usable.. that's all I was trying to point out

and I agree with you, there are a ton of things in proland/ork that are really cool, but they should be thought of more as techniques that *can* work rather than, "hey those shaders look great, how do we get those in unity?" hehe I myself have even said that a time or 2

i'm not trying to be a doubting Thomas about this whole thing either, I sincerely hope he pulls it off. whether he gives it to the community or puts it in the asset store, I for one plan to follow his efforts. I just think 3-6 months is a bit optimistic imo.

 

Cheers for the explanation! It seems with the sort of contiguous terrain i'll be using and where its based, trees that arent very fast are a no anyways, but i'll be running into that quite soon and this seems quite an undertaking so i'll do much as ive done so far in this thread and admire the often surprising things popping up in it and trying to learn a bit behind it, been very educational so far, definitely will be following it, its been a thread that now has me hoping somethings been posted in it for me to very much enjoy

 

Well Im not having much luck with this project. I started this thread about 9 months ago and since then my interests have moved on.

Despsite my best efforts I am finding it increasingly hard to maintain interest in this project and have not made a descend update in many months.

I think most people who where following this thread have long moved on and and I think its about time I did to. I will no longer be making any more posts on this thread.

While I did not get to cover all the features I was hoping to I think I have had a descend crack at it.

Thank to everyone who have added there 2 cents. There have been many bugs I never would have known about if people did not speak up.

The Proland port is still on. If and when I complete it I will make a new thread for it.

Thanks,

Justin.

 

Oh well, it was nice while it lasted, time to learn stuff on my own then

 

Thanks for putting such a great tool out there for everybody!

 

please please continue working on it. I love the project but I can only work on mobile and last time I tried it was very slow for mobile machines. Anything done that would work on mobile would be a life saver because right now my game www.dogfightplay.com has horrible water and i can't find any solution that works fine on mobile and looks good enough

 

Hey Scrawk, did you ever take a crack at porting Eric B's projected grid stuff? I think his z-axis-up thing is tripping me up and I'm missing something in the conversion. I'd very much appreciate it if you could take a look.

 

Hi Scrawk,
I just found this asset in the asset store that is exactly the same as yours (even the UI in the screenshots) and it's selling for 57€:
https://www.assetstore.unity3d.com/#/content/14650

I don't know if it's yours or not (checked the website and it's different from yours, same for the name).
Just to let you know about this.

Edit:
This issue has been solved.

 

Yes, indeed that looks absolutely identical.

 

Thanks. I have seen it and it is not mine. From what I can see it is the exact code from my blog. I have submitted a violation to Unity but nothing has been done. When ever you put something on the internet you lose some control over it and unfortunately some people will take advantage of this. I think as many people as possible should submit a violation so it gets Unity's attention. That way people will not have to pay for something that is free.

I never got the projected grid to work and even Eric's does not work fully (It still lifts at the edges but maybe that is fixable). In Eric's code Z is up so you need to flip the Y and Z axis. I maybe taking another look at the projected grid sometime soon.

 

I did that.

Maybe a good idea to mail legal@unity3d.com or better phone them? https://unity3d.com/company/contact

 

Lol.
From Triton dev.

From the "stolen water dev"

 

Now that's pretty funny, well i guess with ocean renderer not being on asset store he did have to steal the description from elsewhere lol.

Speaking of which, why not put ocean renderer on the asset store (even if for free)? Would kinda avoid those situations i think.

 

Hi all,

Sorry to hear about this.. we are looking into it now..

Best wishes
Caitlyn

 

Look like its been pulled from Asset store, hope no one got caught out and purchased it.

 

It had at least one review. A great one by the way ;-)

 

Wouldn't be surprised it was the dev that stole it who posted to make it look legit.

 

Agree

 

I'd suggest the following to catch this in the future:
- Search for verbatim descriptions.
- Search for matching file names within the submitted package against existing ones, maybe even google (where they're not common).
- Pay extra attention to packages from first time authors, especially if they're technically or artistically complex.

This package could have been picked up on both the matching description and file name search. Scrawk makes use of an EncodeToFloat shader in most of his projects which I imagine was in this package.

 

Because it requires the hassle of putting together the necessary materials to do so, and even for free products you're obligated to provide some form of support.

 

Where do you get that you're obligated to provide support? I don't remember reading that.

 

Most of us don't expect it, but I can only imagine you'd get more emails asking for support wether its free or not. All I'm saying is I can understand why people wouldn't go the extra step from r&d exercise to asset store.

 

That's not quite obligated at all. Also is as simple as not providing a mean of contact to avoid doing support

 

Thank you for your quick response. On a weekend too.

I might have too.

 

I think you should, in all the flavours! I ended up adapting the whitecap ocean a little bit to do much as you said, can now easily lerp between preset water conditions, and with the kind generosity of suimono dev i altered the surface shader to be a little more art than realistic

This is all, however, an impressivebody of work and if you find the time I think you should release all your oceans with a bit of documentation so people can find their way around them a bit, but you could make a beatutiful package anyway, the basic bdrf, phillips and whitecap (im there there's another) would be a very generous asset. My shader writing skills failed to get me a nice shoreline effect but I wonder if you can get shore effect in terms of gentler waves using a map and the vertex ocean shader

I'm gad you showed up, i'll definitely grab them myself if a nice package could come from it, especially with some tidying up, overwater underwater transitions and such

I've played with the whitecaps shader a lot and did notive i had to remove some hdr functions in the shader top play along with Unity's hdr but it's not, at least for me, the resource killer you feared, running at 300+fps on a 580 with the shader significant extended for more arty' control like altering the sky irradiance contribution, add refraction seing through ware and so on. And the fairly-crude-but-osensiby-works method for mixing between sea conditions

You worked a long time on this and impressed the hell out of a lot of people, including me, i'm doing a promo thing soon to help me learn some tools ill be needing soon and the the whitecaps is the ocean tool of choice, this could possibly be finished before you got things up to the unitystore but if it takes a while and you get something up on there i can at least provide a link to the asset rather than just a name/url in the credits