Just a little, at least enough for some glasses/mirror for archviz. Other realtime arch renderers are more flexible about it and allow to handle a certain number of reflective surfaces.
It because UE4 is not arch renderer. It is game engine. I don’t understand what you mean about “Raytraced Reflections”, but I know that Epic still can’t add high resolution cubemap reflections. Not to speak of some more advanced technologies. About this exist thread, for example: Reflection sphere resolution - Feedback & Requests - Epic Developer Community Forums
It is true, it is not an arch renderer but it do it very well. There is not much in the thread you mention, I actually think that cubempas are not the right resource to all the needs of that kind of work. I mean to accurate reflections. Raytracing algorithms have the particularity to give a very good solution to that although they may have an impact on performance for realtime rendering but other applications have shown that this can be kept under control. This is by limiting the amount of reflective surfaces which will be rendered in this way to a certain number or leaving it to the trial/error.
Raytrace technology is very hard to performance. In practice impossible to use this in real time game engine. Maybe in nearest future some changes, but now scarcely. For the time being only voxelized scenes can work with raytrace reflections. If you know some examples real time raytrace reflections, then I will glad to see this. And if you need correct reflections only for mirrors, then try make render to texture material, like this: Mirror shader - Game Development - Epic Developer Community Forums
Take a look at the latest realtime renderers for architectural visualization. Try reflections on them and you will see that it is as I told you, do not expect to find something on youtube. Then re-check the sentence “In practice impossible to use this in real time game engine”. On what theory you base this statement?
Also the scenecaptureactor method performs poorly when it comes to render i.e. a vase on the glass of a table.
Because I’m technical artist in game industry and I know very well how work similar technologies. All reflections in modern realtime applications (though it were a game or realtime archviz) it is or cubemap, or render to texture, or screen space reflections. Nothing more. Of course, exist more advanced reflections technologies, like voxel cone tracing or AHR, but it is very hard to be widespread. Besides, reflections there is very poor and smooth (properly speaking it is reflections of voxels). I advice you learn the gist of the matter and don’t to cherish illusions.
Again:
Take a look at the latest realtime renderers for architectural visualization. Try reflections on them and you will see that it is as I told you, do not expect to find something on youtube. Then re-check the sentence “In practice impossible to use this in real time game engine”.
It is not that I want to discuss with you but I think that you have not idea about what I am talking about when I say architectural renderer. Anyway this is supossed to be feedbak for epic, I don´t understand why you are trying to explain to me your theories.
yes , raytraced reflection is very expensive for calculation , but i think it can be optimize if we have only 1 or 2 depth reflection , in many case its hard to find difference between an image with 3 depth and 5 or 10 depth reflections .
I totally agree. Low depth (=3) and per surface selectivity should work.
It’s almost true you can’t have them in a game engine.
The cue for you is the notion of GAME ENGINE.
As such, while UE is very good in making a pretty render thanks to its phisically based system, it would be impossible to make a game with all the logic going on and raytrace reflections on top of it.
To clarify further, SSR already performs a similar technique, but it’s based on screen deferred data, and can be made fast for practical applications. Of course the tradeoff is losing all the hidden surfaces from those reflections.
The frostbite guys presented a new algorithm that can do contact hardening reflections by raytracing in a pretty smart way, but it’s still a screen space thing.
Bottom line is, even if the UE guys could theoretically program a system that raytraces the scene at quarter res in a coarse way to be integrated with ssr for local sharpness, it would be a major refacoring of the whole rendering system, since raytracing in world space would mean programming data structures to accelerate the trace, all based on mesh data. That’s the stuff for offline renders and realtime renders such as brigade, not general purpose game engines.
They will frankly never make it, since the baseline for UE is PS4 and the whole point of their business is not archviz but games.
Never noticed how they give away the royalties for everything but game projects? Archviz people are doing marketing for them for free in exchange for a good free tool, but it doesn’t generate them any return in cost (at least none that i know of).
That means they will never do big stuff that hasn’t anything to do with games (and by that i mean games that can actually run 1080p 30fps on ps4) since it would be a big waste on internal resources.
Every advance in their rendering pipeline has been driven by their actual internal effort on their games (Paragon and Fortnite), the rest of the renderer is pretty standard stuff in AAA games these days.
I hope i’ve been clear, my english is a mess these days.
Something sounds like wrong to me in your concept of GAME ENGINE, I don´t believe that epic´s people are seeing this just to run something on a mac, a playstation or something even more closed. About the refacoring of the rendering system: then how the nvidia guys have put the vxgi to work over the 4.9? If you are mentioning brigade then you have not the right perpspective about what I am talking, it seems that the matter is much simpler for developers of other software companies. I insist, is just a matter to take a look at the simplest video production targeted arch renderers to realize that it is possible to greatly improve the reflections in ue4.
Also, why the cost should be much greater than the used by the scenecaptureactor method?
Never say never . You’ll be surprised what comes to pass.
Well we must have different perspectives. All I’m saying is that a generalized raytracer to make pretty reflections doesn’t really fit into Epic’s business model, and I say this while looking at the evolution of the engine until now.
Of course we can agree to disagree on this.
VXGI is based on a very coarse approximation of raytracing, and it does so on voxels. It’s by no means a generalized raytracing solution. In any case, if you want very crisp reflections you would need to turn up the voxelization so high that everything would slow down to a crawl.
It’s entirely possible to use just to make offline renders for archviz, provided you can fit all of your scene data into your gfx card ram. Usually VXGI reflections are suited to rough reflections. Just look at the videos and the papers.
Lumion? If so the only extra feature it has are the planar reflections, which you can do (albeit in a confusing and hacky way) with a scene capture actor in UE already. Other techs they have are just fancy names for reflection probes and ssr.
If not, Post your sources then so we can discuss tech over something.
The scene capture actor is hella slow for complex scene, you can do maybe 1-2 surfaces and then you lose fps like there’s no tomorrow.
In any case what i think is: yes, it can be done (most probably not at interactive framerates, i.e. >30fps), no, they will not do it, because it’s too much work for too little return.
They can’t just roll out a feature, they need to make sure it’s useable, stable, performant and generalized enough to fit in the engine workflow, and they need to do this for every platform.
This should not be a technical discussion but a business one.
Just look at what happened to DFGI. That is a feature that has a LOT of traction from everybody, since it does away with so many restrictions and workflow related pains in level design, but they still shelved it in favor of other stuff (performance and VR).
Well, i certainly hope so!
Sure we have and we can have different perspectives,
I know what vxgi is, 5hs ago you was telling me that it was impossible to put any change in the rendering system, that is the thing.
The problem with scenecaptureactor method is surfaces contact. There is more software like lumion, all of them with better capabilities to render refltections than unreal. Yes, performance can be an issue but it can be kept under control. 1 or 2 glass surfaces per room on wich can be placed some objects can be a great improvement for realistic interior rendering.
This is suposed to be just feedbak for epic. It don´t really matter if you think that they are going to do it or not.
Oh c’mon, please take the time to actually read my posts before writing i said stuff i didn’t. Everything can be done, we all know it, but a general purpose raytracer is a ***** to do, not for the technique itself but for its acceleration structures and optimizations (that are very different from how things are handled in engine now).
Again, post your sources so everyone - ESPECIALLY EPIC guys - can understand what you want. I also read another thread by you on this matter, in which you didn’t post your sources. Instead of worrying about raytracing I’d say you should ask “can be <software name>'s reflection system be integrated in unreal? if not, why not?”.
You way of asking is largely unproductive, since from what i gather you’re not a programmer, so just get on with the references.
Also, and this has been stated A LOT by Epic’s own employees on this forum, if any kind of high end technology is not gonna perform on par for at least the baseline tech (30fps ps4) they will probably not bother doing it.
This is a discussion forum man. if you’re not going to discuss anything just write them an email with the request. If instead you would be willing to discuss everyone is still here waiting for actual examples and technical explanations of the tech you would want implemented.