The still images were really impressive (to a layman like me the bar is pretty low but the comments indicate this too). When I watched the video though it broke the realness...not sure how to describe it but everything seemed 'too still' or something. Mostly with the water.
Not like it's less impressive or anything but I was hoping it would feel like the images or something.
To me it at least looks as good if not better than many movie's CGI. It has a little bit of "uncanny valley[0]" but not enough if I suspend my disbelief a little.
That one is a lot better, but still looks off enough that I didn't even approach the uncanny valley. It looks great but it doesn't look real.
In the other video the foliage looks horrible. Come to think of it, I doubt I've ever seen good 3D foliage, let alone examples that come close to real.
the pebbles on the ground looks quite real. But the larger boulders in the background is too textured and not enough bumpiness and also too uniformly flat. That's what seems to make it fake. also the water looks great, but too uniform and perfect. Real water has lots of little flaws and it's usually not as clear too (things like muddy water is hard to render i guess).
This was the video I meant. I agree it looks good but not real. Kind of strange that something can be both when I was expecting it to look real, if that made sense.
What immediately stood out to me was the lack of waves. It's clearly supposed to be an ocean or large sea, but there are only small ripples in the water.
same effect here.
I think I may point to two effect that subliminally distract you from immersing completely in the scene: 1 the water shader mimics moving water but there is no movement on the waterline 2 some aberration caused by the depth of field effect that doesn't match the perceived depth of the elements.
not to detract from it! it's still an extremely impressive demo.
Any of us reading this are old enough to see the progress of video games and the idea of "immersion" happening before our very eyes. We still have faint 8 bit memories and know this is all an illusion.
I really wonder what it's going to be like for people born in the next few years - entering a world where the games are as real as reality. It gives "Second Life" a whole new meaning.
This is very interesting. I never thought about it. Someone who grows up seeing this kind of realism in videos and pictures alongside videos and pictures of real places, and just can't see any difference. Wow...
My kids can't quite believe it when I tell them that we didn't have computers when I was a child.
No internet. 3 TV channels, then 4 when I got a little older.
I'm sort of looking forward to retirement when I can sit in a chair, strap on a VR headset and go spearfishing in Hawaii with my brother. Or swim through space and visit Jupiter, or go on WoW raids.
So. Once the uncanny valley gets crossed, and we can successfully create synthetic images - and eventually videos - that purport to show something happening that never actually happened.... will pictures and videos no longer be valid evidence in court because they could be made synthetically in a video game engine?
Going further: could wars be started because of something someone made in a video game engine?
believable fakes have been produced with photoshop for ages. raytracing has been able to produce photorealistic images for decades (although not in realtime, but that doesn't matter here).
A random photo or video with an untraceable source already won't be treated as serious evidence in court.
A video + a testimony "This is what I saw on that fateful night" or "I retreived this video from the officer's body cam immediately after the incident, it was in the custody the whole time and forensic analysis shows no sign of tampering" carries an entirely different weight than that video alone.
Regarding wars, in the recent conflicts we already have cases where doctored photo & video evidence was used as arguments and vice versa, where real evidence is dismissed because it can be made synthetically - and it obviously can, they become evidence only if you[r organization] gathered it yourself or you really trust the soruce.
Wars definitely won't be started because of something made in a video game, but such videos can and most likely will be used in justifying a war decision to the global public and building the relevant PR/propaganda materials.
I haven't played any 3d video games in about 15 years. The video at the bottom of the page looked awesome, except for the edges of the water. I guess when they do Engine 5 they will finally change the name to 'Real Engine'.
I don't mean to take away from it needlessly or criticize gratuitously, but if you watch the video you can see the water turbulence is still unrealistic. I guess this is a very difficult problem for the unreal engineers to solve because the physics are hard.
I never know if "physics are hard" or maybe they already have working models that are completely realistic, just impractical to run on "normal" computers.
So UE4 is capable of rending still images as good as a standard 3D image rendering program? That's nice but until these are rendered at game level FPS it's not very meaningful for games.
A question for those in-the-know about the state of VR: Could a home desktop render these scenes in high-enough resolution in RT for the upcoming headsets?
I think in part it's down to the fact that we want to interact with water - interactions with fire tend to be more incidental, and therefore there's less incentive to simulate or mock it up to a higher standard.
Something I've been meaning to find out for a while: we've all played with tearable cloth simulators in Javascript or HTML5. Surely game engines are more powerful than these, yet even modern games have cloth that doesn't move, stops bullets and can be stood on like it's concrete. Is it just not a priority/not worth the cycles?
It is less than easy to interface something like that with a full game engine. Things like requiring collision culling (not doing the whole O(n^2) "check if A is colliding with B" thing), doing proper shadows, being properly deterministic, being alright with variable framerates, that sort of thing.
Also, I don't know what simulators you've seen, but the ones I've seen take an absurd amount of CPU time. Or occasionally GPU time. Works well for a proof-of-concept. But for a full game... Their priorities are elsewhere. Not worth the cycles.
That being said, I fully agree that game focus should be shifted slightly away from straight graphics and more into gameplay.
I don't work in 3D graphics but it seems challenging to me because of the fluid dynamics. Fire is very low density gas rushing upward into the moderate density air - and even if you can do all the fluid dynamic calculations to render the boundary of the flame, you also have to render the light it's throwing onto the surrounding surfaces. I have no idea how close technology can come at this moment but I'm not surprised the answer is "not close enough for prime time".
Wonder how long we will be stuck in a 99.X% era of realism. Our brains still picks up the small things and tips us off that something isn't right. Especially when things start to move about.
Not that it's not cool to be on this journey. Every one of these iteration towards realism gets me excited.
The problem with "realistic" graphics is that they only look that good in screenshots, and maybe a simple video. But now add a person in there. Then show them talking, walking on the sand and through the water. Is it still going to be realistic? Not likely. To the people who are saying we are getting over the uncanny valley, I think you are wrong. I think it only gets worse as the environments look more real but the physics and characters don't match. When everything looks fake it's easy for your brain to accept, when it looks real but doesn't behave like reality it gets weird.
This is all coming from someone who is still playing video games from several generations ago, so you can just ignore me if you want to.
This is mostly due to the fact that they use models and materials that are easy to render. It's the things like flesh and plants that often stand out, either because their poly count is too low so they don't look like they should, or because the material rendering isn't there yet.
As always, animation is definitely the hardest thing to get right when it comes to 3D. The best animations out there still look unrealistic, it must be incredibly frustrating for animators.
Often the most impressive screenshots are the ones that look least impressive.
It's (comparatively) easy to have screenshots that look great if it's using a lot of assets with baked lighting. It's a testament to the designer's effort, but says very little about the engine.
Real-time ambient occlusion, radiosity, global illumination, refraction, etc... that's the kind of thing that is (technically) impressive in an engine, but so var very little of this (beautiful) Unreal/Unity/* demos actually deliver that. And that's a shame.
The craft deserves praise, but there's a distinction between asset creation praise (to the designer) and technical rendering praise (the engine).
One reason these look so good is that they are making good use of depth of field. It isn't something typically used to this degree in games (it would be a hindrance to play with).
Beyond that, depth of field typically isn't carried through into reflections (you can focus your eyes on marble floor or you can focus on the reflection in the marble floor).
On the best looking beach shots, I don't know if the unreal engine blur the reflections to make the depth of field look better or (more likely) the artist saw the reflections as too sharp and softened them, which looks better because they would be affected by the depth of field.
71 comments
[ 3.0 ms ] story [ 129 ms ] threadhttp://www.simonstalenhag.se/
And just like someone else pointed out, it does remind me of Simon Stålenhag. I even own one of his prints. http://www.simonstalenhag.se/bilderbig/fjarrhandske_1920_bad...
[add] actually, quoting the author: "lighting is really basic, movable direct and sky, without any GI/DFAO/bake/"
https://youtu.be/_nLGoqqDc0w
[1] https://forums.unrealengine.com/showthread.php?58385-Koola-s...
[2] http://tof.canardpc.com/view/3934ce8b-4423-4dd6-a16f-3e35376...
To me it at least looks as good if not better than many movie's CGI. It has a little bit of "uncanny valley[0]" but not enough if I suspend my disbelief a little.
[0] https://en.wikipedia.org/wiki/Uncanny_valley
In the other video the foliage looks horrible. Come to think of it, I doubt I've ever seen good 3D foliage, let alone examples that come close to real.
https://forums.unrealengine.com/showthread.php?58385-Koola-s...
not to detract from it! it's still an extremely impressive demo.
I really wonder what it's going to be like for people born in the next few years - entering a world where the games are as real as reality. It gives "Second Life" a whole new meaning.
10 years time: "When my daddy was small like me everything was made of pixels"
No internet. 3 TV channels, then 4 when I got a little older.
I'm sort of looking forward to retirement when I can sit in a chair, strap on a VR headset and go spearfishing in Hawaii with my brother. Or swim through space and visit Jupiter, or go on WoW raids.
Going further: could wars be started because of something someone made in a video game engine?
in my opinion, this doesn't change much.
A video + a testimony "This is what I saw on that fateful night" or "I retreived this video from the officer's body cam immediately after the incident, it was in the custody the whole time and forensic analysis shows no sign of tampering" carries an entirely different weight than that video alone.
Regarding wars, in the recent conflicts we already have cases where doctored photo & video evidence was used as arguments and vice versa, where real evidence is dismissed because it can be made synthetically - and it obviously can, they become evidence only if you[r organization] gathered it yourself or you really trust the soruce.
Wars definitely won't be started because of something made in a video game, but such videos can and most likely will be used in justifying a war decision to the global public and building the relevant PR/propaganda materials.
https://forums.unrealengine.com/showthread.php?58385-Koola-s...
I don't mean to take away from it needlessly or criticize gratuitously, but if you watch the video you can see the water turbulence is still unrealistic. I guess this is a very difficult problem for the unreal engineers to solve because the physics are hard.
It is less than easy to interface something like that with a full game engine. Things like requiring collision culling (not doing the whole O(n^2) "check if A is colliding with B" thing), doing proper shadows, being properly deterministic, being alright with variable framerates, that sort of thing.
Also, I don't know what simulators you've seen, but the ones I've seen take an absurd amount of CPU time. Or occasionally GPU time. Works well for a proof-of-concept. But for a full game... Their priorities are elsewhere. Not worth the cycles.
That being said, I fully agree that game focus should be shifted slightly away from straight graphics and more into gameplay.
Not that it's not cool to be on this journey. Every one of these iteration towards realism gets me excited.
This is all coming from someone who is still playing video games from several generations ago, so you can just ignore me if you want to.
https://www.youtube.com/watch?v=rbDQYXNDXKM
It's (comparatively) easy to have screenshots that look great if it's using a lot of assets with baked lighting. It's a testament to the designer's effort, but says very little about the engine.
Real-time ambient occlusion, radiosity, global illumination, refraction, etc... that's the kind of thing that is (technically) impressive in an engine, but so var very little of this (beautiful) Unreal/Unity/* demos actually deliver that. And that's a shame.
The craft deserves praise, but there's a distinction between asset creation praise (to the designer) and technical rendering praise (the engine).
Beyond that, depth of field typically isn't carried through into reflections (you can focus your eyes on marble floor or you can focus on the reflection in the marble floor).
On the best looking beach shots, I don't know if the unreal engine blur the reflections to make the depth of field look better or (more likely) the artist saw the reflections as too sharp and softened them, which looks better because they would be affected by the depth of field.