pouët.net

So, I was just watching some of this guys videos.
Apparently these are done using the Unreal Engine 4, and run in realtime at about 30 fps on high end hardware.
Now, things like this, this and this would make a damn nice demo, I mean slap on some text overlays with scene poetry and we have a compo-winner! :D

Since I am more of an oldschool coder (haven't done any PC/3D stuff since 2002) I probably am asking stupid questions, but still:

1. Could rendering of this quality be done by a "scenegroup"-coded engine, or does it really on heavy-duty-large-team-commercial-only-codebase massive algorithms for illumination models etc.?
2. Or is the hard part content creation, ie. designing objects/textures/lighting?
3. Or is that noone really cares about "realistic"-looking 3D flyby demos anymore? :)

I'd say a lot of this does not only depend on used rendering techniques and the engine, but a whole lot on creating good-looking assets. And creating those takes a lot of time which many people do not want to invest into a demo.

It is number 3.

2 & 3, it seems to me. However,

I WANT THOSE IN A VR HEADSET AND TO WALK AROUND IN THEM. *ahem* Sorry. They were gorgeous. The leaves could have used a little more work in the second one, and I don't know why the water in the first one still seemed a tad "not quite". The third one, wow. :)

Anyway, carry on. Again, sorry for the caps.

Speak for yourself, Navis.

mmh. 1) too natural. 2) too polygonal. 3) too cheap. it's colors but just lightmass and a reflective floor. but... who am i to talk?!?

Top demosceners could definitely make something like that, some of them work for game companies as their daily jobs anyway. There is nothing really special going on, most game rendering tech is public knowledge.

Definitely takes a lot of time to make high quality assets. The current trend in high end games is scanning, photogrammetry, motion capture, etc... Given enough time it can all be done but I would say most demosceners with real lives, family and fulltime jobs don't want to spend so much time to make something for free that will be seen by maybe a few thousand people and then get a thumb down because someone didn't like the colour choice.

This is personal opinion but I think flybys of pre-generated static scenes isn't very demoish. The best demos to me will be the ones with original ideas, great dynamics and slick design combined with great music and clever coding. Demos have their own style and are impressive for different reasons rather than trying to copy cutscenes from games. I hope camera panning around high definition scans of rocks is not the future of demos.

it's the old art (modelling, textures) vs technology (shaders, shadows, deforms, physics) thing, I like 'effects' a lot more than flybys, but I love demos that stage their effects in a proper 3D scene too, with modern lighting methods usually

praise cocoon, fairlight, mfx and some others for putting extra effort in modelling and texturing objects and environments that fit the flow and theme instead of being yet another parking garage or alien lab

and those unreal demos are stunning, all of them, not gonna lie

no because

Its the assets. Most of 3D guys are already working on heavy stuff, less time to do free stuff.

1) yes, realism is mostly the right combo of good shaders (plenty of prefab shaders are available in ue4, so less effort is needed on that department) and good assets
2) preparing good shaders and good assets (and combining/tweaking them in your own engine) is a lot of work and a team effort. also, the number of 3D artists in the demoscene is quite low and the 3D you commonly see is skills wise not super high-level either (and i happily contribute to that as well, it's a hobby, not a job :P).
3) plain flybys are deemed boring usually by the general demoscene audience because they lack 'effects' in a classical sense.

<wild prediction>
Actually, the current trend towards Physically Based Rendering could help (and actually already does if you look at eg. Mercury prods) the 4/8/64K categories a lot. The shaders and potentially the pipeline get somewhat more complicated but that's not a lot of code in the end. Plus, procedurally generated textures and materials suddenly look good again.

When we're talking about full-size demo engines, in the end it's "even more work". I can understand that most people rather limit themselves to more stalized looks because they're easier to pull off when you have a life and artistically more interesting, plus you avoid people saying "I can do that in 30 minutes in Unity" :D
</wild prediction>

(Also I heard that TBL are working on a full port of Frostbite to Amiga so we're all screwed anyway)

No, I don't think this level can be achieved by any demo group.

There is an absolutely incredible metric fuckton amount of work put into UE4 engine, just look at the demo with a careful eye. We are talking about thousands of man-year amounts of work here, minimum, and that work was done by hands down the very best guys in the industry. And it's not just the rendering, the asset creation chain is at least that important, if not more important. Probably even more important.

No fucking way a 5-man strong demogroup in their free time can do anything like that.

Thats what they think, they can do. Reality mostly looks different. ;o)


:-O

blala, then again a generic game engine might need some things that a demo can do without.

about 60MB of things in fact!

Surely the thing with demos is the fact they are pretty much hand coded from scratch, essentially proving that coders can produce something absolutely unique in some occasions.

The idea of using a game engine to produce demos just seems a tad "lazy", that coders aren't interested in showing their coding prowess.

blala: Yes fucking way. The work done on UE4 is almost exclusively the infrastructure. The rendering is maybe 5% of that.[citation needed] For a demo, you can get away with a small fraction of that again because you can restrict yourself to what you want to show. Your stuff works only when shadows are not important? Your IBL looks like shit except in outdoor scenes? Your reflections suck on higher roughness levels? Your antialiasing breaks down in dynamic scenes? Not so much of a problem if you control all the content and camera angles. Use sleight-of-hand strobe flashes and lens flares and text overlays or whatever fits your style to hide the rest of your problems. A game engine can't do any of that. Instead, it has to spend a huge amount of complexity to make all the edge cases work because people pay for not being surprised during development of their game. And even after that huge amount of work on UE4 rendering by incredibly good people, there's still some cases in which things look really bad. (This is also the reason why those demos look better than the games will. You only see the cherry-picked best parts, like in a demoscene production... Screen space reflections sometimes look like shit in the bathroom mirror? Don't show them. Oh wait.)

It sure is a lot of work, but getting close is manageable. Being better in certain areas is possible as well if you specialize. And since restrictions fuel creativity this is not a bad thing at all.

realistic looking 3D-flyby demopart (atleast it was back in 93!):

Fact: UnrealEngine is being developed on by a very big team (even bigger than Titan!) for ~12 years by now!

1.) Sure, if the Group has many coders and they all either are jobLess or can reUse the Code they coded at their JobTime (the last years) !
2.) Groups normally have enough people to spit out enough content for 2-5 minutes!
3.) correct!

It´s all about PBR...once you got it right, you can let it look alike what UE4 delivers pretty easily! While my winning-Revision-8k doesnt make any use of PBR (it´s DirectX9 still :p ), i still may have a tipp for you, simply download this here and build on it...or use as is! ;)
Get there, click the "Downloads (YourSystem)"-Button on the newest version, choose "Built In Shaders" and get your own PBR done without coding a single line, just by deleting everything unnecessary! ;) (can take ages, i guess!) ...StandardShaderCore could be a good starting-point, maybe! Just don´t get lost in there! ;)
http://unity3d.com/get-unity/download/archive
(i atleast could easily extend this to be able to crossfade between two textures in "30 minutes", so it should be possible to do whatever you want with it! ;) )

Fact: "Until mid-2008, development was exclusively done by Tim Sweeney, founder and technical director of Epic Games." [1]

Meaning, an Engine that was like "the-best-around" was developed by a big team for 5 years, then got maintained/expanded by a single person? HEAVY!
...but i meant all around it aswell, like Editor/Examples/CustomerSupport (feeding Tim with good Ideas to work on, etc.)
...so basically i mixed 2 things up..."Engine" and "SoftwareDevKit" ;)

ALL hail UnrealTim!

Err did you read the article?

Actual quote:

1- possible with lots of work and skill :=)
2 - lots of work and skill to do "next gen" asset ^^
3 - flyby is an art :o)

with full time job and money all is possible :)