[Coming Soon] SpectraGI Advanced Real Time Global Illumination For Unity from Livenda

https://research.nvidia.com/publica...ctrees-analysis-extensions-and-implementation
If you are interested in rendering programming, It`s worth checking out.
The paper has a detailed description on how SVOGI works. It even has source codes.
They aren`t that long so you can have your own SVOGI renderer implemented by tweaking the algorithms a bit.
Understanding the basic algorithms and how the entire system works, although not insanely difficult, can be quite daunting. But it`s worth it.
As long as you have some basic understaindg of linear algebra, physics, and calculus, you are good to go.
If you like physics and math, it`s like a goldmine. Not only it will help improve your math, C++, and physics knowledge, you get to get a glimpse of how the rendering system interacts with windows` graphics API.
For me, the most difficult part was low-level stuff(handling errors and memory issues,,)

If you are serious about programming related to making games(not some mediocre side-scroll platformers), I think it can be a good starting point.

 

UE4`s dynamic weather plugin.
Looks like UE appstore is getting bigger every month.

 

That asset is in unity as well

 

I wonder how heavy the GI in Unreal 4 is, does it have a big impact in performance if you need it to be very dynamic ?

 

By default, there's no realtime GI (just lightmapping). You can enable light propagation volumes (LPV), and that's pretty resource intensive.

 

Very interesting, i could program the GI Proxy method for Unreal store as well later. I will look into into it.

Hopefully they handle many point lights well

 

You could. Actually UE is very good at handling a large number of point lights.

 

I know. But the point is, with the source code open to the public, the quality of the apps which perform much better, tightly connected to the engine source code without any additional layers(wrapper class,,etc) will be better in the long run as the devs have direct access to the engine source code.

 

Great, would be a perfect fit

 

Hello Everyone,

Just a quick update, we are working vigorously to complete SpectraGI (not just the videos) which is after all the most important part We know everyone is waiting to see it in action and other details, rest assured it will be unveiled soon, as soon as it is ready.

Important: We kindly ask people to refrain from posting and discussing topics and issues unrelated to SpectraGI. This thread is not about general Global illumination and other products, UE4 and other unrelated topics (you are more then welcome to start your own thread on a relevant forum section if you like to discuss anything else) . Also, there is no reason to compare Enlighten at this stage as it is irrelevant and we are well aware of the differences with SpectraGI. There are a myriad of technical differences (for example, almost instant pre-process, handle very complex scenes with 'thin' walls with NO bleed & UV issues & UV dependency, SpecraGI can also natively use normal map data for directionality for diffuse indirect irradiance to bring out detail without directional lightmaps and much much more ) which will become more apparent when we release the details.

We cannot wait to show what we have so far and complete SpectraGI for availability as soon as possible.

Cheers!
Livenda

 

Have I somehow missed the second area lights with shadow demonstration? 8-/ . Would really like to see the shadow quality. I´m using shadow softener at the moment which is really quite good, but I´m not sure if it will work with SpectraGI..

EDIT: Gosh, I was on holiday for the last week and was totally disconnected from the internet. I just scanned through the latest post and found.. nothing new. Livenda, your products may be first class, but your marketing is awful (I know, I´m repeating myself :-/ ) . Really guys, enough with these teasers. Post about the development, the problems and things like that. We´re not daft..we can understand technical stuff .

EDIT 2: My god, I counted 3 (in words THREE) "soon" in the last post ;-) .

 

To paraphrase Sir Hammerlock,
"Apologies, but when ALUNBRA speaks, I feel my brain cells committing suicide, one by one"

But I digress,

You might be aware but we are not, if you create a void of silence after making really quite bold claims and pre announcement announcements of what you might be starting to do, then you will get a flurry of people talking amongst themselves while they are waiting. I'd be much happier talking about your product more if only you would, I have no problem with you taking time to perfect it and I appreciate the complexity, just expect a thread full of people guessing and nattering if all you do is tease

Nothing but words on the subject still though, reminds of this old quote:

"Writing About Music is Like Dancing About Architecture"

Pictures Damnit!

 

damn... another coming soon message. "Coming soon!!! more coming soon talk!"

 

lets not talk till videos

 

Pretty sure the one singular thing keeping this thread alive right now is hype.

 

You are lucky, some of us have been hiting the refresh button here for weeks, i still remember that friday when the GI video was encoded and coming.

 

oh i can keep quiet for 24 hours - it's ok or even or 24 days...
i can even stop sleeping, anyway my nights will be lit by SpectraGI when gets released Someday ...

 

Mastering the patience is a big challenge. I seem to have mastered it now when i don't have my refresh key anymore...at least what was written on it, after the first two months from releasing that video

 

, you got in the game early, that is why you have reached that state

Btw, i love your collider free raycast system, could be very usefull in many cases i had problems in both my particle and GI systems, really cool work, i only recently found out about it

 

Maybe we will all 'soon' be a part of history here, after this is released, it will become the GI method for everyone who wants fast and accurate GI! And when our grandchildren talk about how amazing their graphics are, we can tell them that their grandpappi was there when it happened...

(maybe even pull up this thread if they don't believe it!)

 

yeah when i see "GI" mentioned in any thread every time i hope that i will see that quality that half of the population of the world waits to see in the 3d world(like most of you guys). I always loved the idea of having great physically lighting simulation in 3d space. and because everything is about the raycasting/raytracing - that led to creating of the DRCS plugin.
I won't ever forget the moment i saw the little strange bird shaded so good that i was wondering for a long time what for a guy can paint such a great looking 3d bird model. It was about the time of the nineties when I heard about "3ds max" and the GlobalI llumination. So i believe in a few years we are going to play games which will have pretty realistic look. Everything is about the raytracing as you can see in Brigade Engine 3(Sorry LIVENDA for mentioning Brigade 3 i'm sure Specrta will beat every GI solution out there and will satisfy most of us )

Cheers!

 

Hopefully raytracing will be the next big thing, there is already some cool showcases of real time raytraced scenes with noise artifacts in motion due to pixels not drawn fast emough (i guess), so the whole thing is not that far away, at least for some of the effects.

This is what we are all waiting for since the first days of 3D gaming, i am sure

 

I think the next big thing is going to be hybrid raytracing built directly into gaming GPUs, since shaders/polygons are still great for hard surface shading, and hybrid raytracing picks up the slack for transparent surfaces (for getting rid of transparency artifacts once and for all), reflections (no more cubemaps!), and shadows (pixel perfect shadows, anyone?)
Imagine's already got cards with raytracing modules in them, although what we really need is for ATI and NVidia to pick up the technology.

 

My opinion is that Raytracing was and will be the perfect lighting solution for a long time. As you probably know the raytracing gives you naturally most of the effects you will ever need such as reflections, refraction, motion blur, DOF etc... The only reason we don't have RT lighting solution based on pure Ray Tracing is the hardware performance standards because the GI algorithms that are known are already 40 years old and are not changed till today..... Please nVIDIA and AMD make some monster GPU 2 time faster than Titans or R9 at the half price

 

I agree,as soon as I saw their demo I knew it was the future,can't wait until I don't have to look at shadows getting sharper /less sharp as I move closer/further away from them!

 

I also think that could be something that can provide really good looking results but only if it's hardware accelerated as you say. I know Imagination Technologies - PowerVR SDK has some kind of Hybrid support if i'm not wrong but is still slow... The only thing i can imagine that's able to provide RT GI performance is some Screen Space solution as the GPU can do everything needed to calculate GI and AO. That's why i believe SpectraGI will do it with the cost of many hacks and workaround. I only guess
Go LIVENDA.

 

Sorry of missing that . But you will really need the hardware to see it in RT. I don't see nVIDIA or AMD go into this but who knows .
One thing i'm concerned... When we are going to move to DX11 and OpenGL 4.3-5. Even though DX9 is the most used API for gaming it's to old. DX11 and DX12 used in XBOX ONE as OPENGL used in PS3 and PS4 have now many advantages that can boost the graphics quality a lot. We just have to start build games on that (old)new platforms to get better quality. IT IS time because even 5 years old hardware is DX11 and OpenGL4.x capable...

 

This is exactly what PowerVR is; a hybrid, and Unity is integrating it at some point in 5's life cycle, and runs on mobile, obviously we cannot do full multi bounce GI in realtime with it.. yet, basically Enlighten will get you some realtime ish results on low end to mid, SpectraGI will be realtime for high end if it works ;-) and PowerVR will be what we use soon for realtime shadows and reflections, and then in the future to brute force GI on mobile one day, I think a bit like we use Nvidia PhysX for physics PowerVR will be for raytracing, also Enlighten are plotting moving to GPU at some point and removing the pre calculation could be the result, so lots of options in Unity going forward.... and we're off topic again :-D

 

you're right but what else we can talk about Spectra? everyone wants to know more about it but no one does
Sorry i know everything but am not allowed to share.....

 

So should we place bets for video announcements and then the final GI release ?

Area Light Shadow Video: 1-2 weeks
Realtime GI Video: 3-4 weeks
Realtime GI Release: 6-8 weeks

 

Helios why do you think that Spectra will only work on high end? Is it becouse of number of light bounces or basing on being fully dynamic?

 

The fact it's explicitly stating directx9 support counts a little against the high end stipulation

 

Or the opposite, since you would need a stronger system to run the DX9 effect, than if it took advantage of the GPU fully through DX11

Unfortunately could be many months until we have an answer to that.

 

I find that kind of idea a little backwards though, as some things you can do with dx11 actively trounce dx9 equivalents and render it somewhat redundant, which is why the question was raised regarding how you could feasibly reach certain results simply in fragment shaders over DirectCompute

 

In any case though, the system would need a beefier DX11 system than it could use in a DX11 mode, or perform worse on a lower end DX11 system.

Personally i love the fact that it is DX9, since i dont use DX11 , so if it is optimized enough i would use it in my game when there is no battle or very demanding scenes.

I am sure it will need a beast of a DX9 system to have any chance to run in full real time in a complex scene that is already on the edge of 30fps though on a mid range machine.

I dont know the exact technique they use though, their screen space refrections are very fast and defintly a very good sign. I guess that is why we are all still here , expecting them to perform a miracle.

 

I don't even develop full time and I just cannot justify including an feature that affects the entire art style and direction if it takes such a huge bite out of performance. It's simply not worth it, there are many, many things going on in my proposed projects which span a variety of fields and the sort of hit you're talking about is simply untenable, even in the field of speculative prototyping. You can't get people excited about a feature if the product is a skeleton upon which the feature hangs off and you can't entertain any feature at all when the insight into its development is as opaque and if i'm being positive, badly handled and if negative, manipulative as this

Every major asset i've used regularly i've had communication with the creator, we've discussed things, sometimes its not been the right time, sometimes its great for what i'm doing, i feel generally the asset will always be there and i'll always give them a bump when I can, i've really enjoyed this communication and transparency and that the asset creator is a professional but, like his/her customers, an enthusiast who is willing to get involved with their customers creations, this is good support, this is what makes great assets, anyone can PM me and i can give you a list of assets i can absoutely assure you have stellar support and they really do care what youre doing with them (Although they roll their eyes at me at times im sure). This nebulous shroud around this thing s all very glitzy and shiny aftereffects but to be honest, I own their previous asset and the demo scenes weren't even completed to what I considered a polished standard. Have the new features proposed been added?

Didn't this exact same thing happen last time when their previous product was announced?

It's hard to just be quiet about this stuff, this post could be deleted, but it would be just censoring any valid criticism of a product that will go onto the asset store and sell a giant load as it fills a hole (although any gpu whizz and last few editions of gpu pro should have been on this issue by now), but the way it's being sold is really not fitting with my idea of good asset advertising or support

 

Your talking about the "support" of a product that doesnt even exist to the public yet. Criticise support after the product is released... because right now there is nothing to support. I think your issue is lack of transparency. Thing that makes people nervous, or at least me, is those "when its done" comments. Anytime I hear a developer say "when its done"... it tends to be vaporware or they have some significant issues. Which makes me nervous even more because performance is a big deal.. even if they can get it looking pretty it has to run pretty.

 

What support means is entirely semantic and you can have your definition and i can have mine and i tend to take support as from where the asset is first announced, the developer creates a communication, they take suggestions and feedback, they respond to a group of people who are intelligent and informed and both sides can learn from each other. This is support. Then afterwards I would hope for a similar level of interest. It's worth noting that my support for, again, i wonder what product, amounted to 'well actually they should make their product work with ours', when i'd reasonable enough information that the slightest effort might have been made. Look at the Skyshop or RTP or Lux(Which is free!) or god, any number of threads really, great work by real professionals who respond and try to integrate and adapt however they can. Not '*shiny aftereffects video* see you next time suckas'

 

I will say the original after fx video was retarded... u going to have a minute or 2 of text and then 2 screenshots... yea... u could have just wrote that in a post... lol...

 

the deed was done and faded into nothing.
,,same can be said for this asset.
why is it so hard to make a video about it?
Rather than saying "as soon as it is ready", post a 10 second-long video to prove skeptics like me wrong.
no one's gonna blame you for the quality of the product. we just want to see if it actually exists.

 

DX9 doesn't mean much of anything, they have to support OpenGL whatever they do and if it supports OpenGL then it can be done without DX10,11,12 like features

I'm sure it will be relatively high end because there is no fast way to do GI unless you pre cache and pre bake parts of it like Enlighten does.. for now.

If it supports low end it will be at lower resolutions.

If they came out with tech that was high quality on low end devices and fully realtime I would be the single most amazed person the world has ever seen, As any good scientist I won't say it is impossible, but it is however quite improbable.

In any case there is plenty of room for a product like SpectraGI may? possibly? likely? be? one day? soon?

 

This. If there's a way besides Enlighten to do fast realtime GI on low-end systems, I'd have thought Epic or Crytek would have jumped on it long before Livenda.
Of course, I could end up eating those words, but I don't think so.

 

That's the crux of it isn't it. What they're claiming simply sounds too good to be true and yet we all hope against hope that it's real. The problem is we've seen practically nothing to back this up and repeatedly assuring us that videos are coming — videos that never eventuate — doesn't bode well.

I find it hard to understand why a video of an almost ready product is such a big problem, surely if the product will be unveiled soon, it must be nearing the end of it's development and therefore be capable of coughing up a demo of it's capacity...

... and like everyone else following this, I'd be happy to have my doubts squashed with a demo.

 

When they say they will release it soon, it means they will release it soon! They don't need to be reminded every 6 months about it ; )

 

^Hahahaha...

 

GI is trully epic, the wait is definitly worth it

I tweaked the GI in my approximation system in the castle scene i am currently working on and the result is unbelievable, all the detail that was hidden before is now evident and everything take a sweet extra coloration, it is a marvel to behold even in the non exact solution i use.

Now i have to put in all my other maps (100+) and combine it with the IBL addon i programmed just yesterday, hopefully Spectra is also out by the time i am done so i can implement it too for even more subtle results (or Enlighten anyway, whatever works best in combination).

 

Well, I´m slowly losing faith in this over-hyped product. . It´s very annoying that we don´t know what´s going on and can´t rely on their statements.

 

If I had a working GI system,
1. I would post regular updates on how everything is going.
By updates, I mean "we are trying to resolve deferred rendering-related issues or something like that
not "Just wait, we are almost finished".
2. Keep people up to date on the quality of art work that it is capable of producing.
3. Silence skeptics, including me for good by showing a video or two with fps checker on.

If you had a real GI solution, I highly suggest you to do one of the above.

Stop giving people with less rendering-tech related knowledge false hope that they can achieve V-ray quality real-time
game with instant pre-processing time. There`s reason why people with some degree of rendering tech knowledge remain skeptical about the validity of your past statements.
I`m no expert. But I know enough to understand how rendering part of the engine utilizes the graphics API.

Either you have come up with some new revolutionary mathematical algorithms specifically tailored to GI that can shorten the amount of the calculations needed to be done or greatly approximate the process, or extremely efficient parrelel rendering algorithms that work with any kind of shader(which, to my knowledge, is highly improbable) or blazing fast noise filtering algorithms to reduce the noised caused by ray tracing.

The possibilities are limitless. Show us something.
And stay truthful to your words.

 

http://on-demand.gputechconf.com/gtc/20 ... n-gpus.pdf
Rumor has it that this year in Q4, major engines are going to implement the new GI method called VXGI.
No cheap screen space tricks, or pre-baking.
Basically it`s advanced SVOGI that works well on modern GPUs.
It uses Clipmaps instead of mipmaps. Clipmap requires no nodes or pointers

http://money.cnn.com/news/newsfeeds/articles/marketwire/1145759.htm
It will be integreated into engines like UE4, Cryengine this year.

This is the nvidia video showcasing the quality of their GI.
It looks amazing...

We can now give the middle finger to lightmap-baking.
With major engine companies jumping in on this tech, it is most likely not just a hyped solution.
Every in-game object, including skinned mesh and moving object, is affected by indirect lighting.
So no static objects,, which is just mind-boggling.

I feel bad for Unity`s enlightn integration,, It requires ridiculously huge baking time for good results,,
And we, as indie devs, don`t have that much budget.

If you are curious, read the pdf file linked above. Pretty exciting stuff.
Hopefully they will publish papers on this so I can get my hands on it soon,,

 

This is the method Livenda uses ?

Also i think voxel techniques have many issues, like not grabing smaller or animated objects etc, so they are far from idea solution. To integrate this in Unity would have to find specific ways to voxelize a complex scene and handle the in motion/animated items by keeping voxel grids for every move, or by animating the voxels (which i think defeats the purpose a bit since would cut back on speed).

Seeing a static scene that i could emulate in my GI Proxy system in a few minutes and not be able to tell the difference does not mean much, plus GI Proxy works in Unity free and is a 100% dynamic solution grabing all animated objects of any size and detail and works without any baking.

These systems are all tested when motion and heavy detail kicks in, until we see this spaceman multiplied x10-x20 that will also start running around and the sun change/move in a combat scene with AI and a thousand other things going on, with good frame rates, this will just be another unusable in an actual game technique.