I love the quote at 4:08 "A lot of professionals on the internet said our claims were impossible, but they were wrong and now our technology is being used by some of the world's biggest companies!"
It's like they understand how shady they look and sound, but then say "It's alright, don't worry about it!"
Bruce dell is the ultimate troll, he says these things to ride the wave of angered people. Unfortunately for him the exposure is for naught as the actual product being sold isn't good.
And of course scan data fits every level design... gee, it might be a good tech for google street view or something but it's totaly not aimed at the right crowd...
I wish they'd stop talking about games. The lighting is static, nothing is dynamic, no spec ... nothing. It's basically one large model. It's impressively detailed but that's about it. They've got maybe 0.001% of what is actually required to make a game completed.
they still have a long long way to go to make it into games, and I honestly don't see it happening. the environments they showed looked like current camera projection techniques combined with of course the nice high resolution model detail scanned. Other than that, no reflection/spec and dynamic lighting really made the images look very dull.
Its incredibly detailed geometry rendered full-bright. Are they claiming this is real-time lighting? From what I can tell they're claiming their models are very detailed, which the are. The rest is capture of a photo. They haven't come out and said specifically thats what they're doing, but I haven't seem them claim the alternative either?
Its incredibly detailed geometry rendered full-bright. Are they claiming this is real-time lighting? From what I can tell they're claiming their models are very detailed, which the are. The rest is capture of a photo. They haven't come out and said specifically thats what they're doing, but I haven't seem them claim the alternative either?
Avoiding to answer direct questions and any criticism due to wanting more investors.
They _are_ claiming that this is actually better than the tech that is currently out there, as they have been claiming for like 10 years.
I wish they'd stop talking about games. The lighting is static, nothing is dynamic, no spec ... nothing. It's basically one large model. It's impressively detailed but that's about it. They've got maybe 0.001% of what is actually required to make a game completed.
Right, the tech is very cool for certain applications, like engineering, architecture vis, 3d walktroughs (would be cool for the real-estate industry), but it still is a really poor fit for games.
I mean if their Software can do all that...i mean why not make it an Warch Viz app? Why try to get into the Game business, that is populated with Software that have stablished Workflows and ofcourse need 100 people teams (eventhough that isn't even true) because what is the thing for this generation is interactivity and sandbox not 100% rebuilding to the cut in the wood and thats it. I mean current games make a good enough case for visuals right now.
I just stopped the video at certain points of the claimed real life (That looks nothing like real life) and I actually realized how unrealistic it looked like even worse than I thought.
Half life 2:
Their claims of "Real Life" (Give some pity because of the compression)
Game from 2004 looks better in my opinion
It looks like a environment from someone fairly new to 3d that was dumped into sketchfab or something with little care or effort.
I couldn't believe it...HE TOTALLY SUPRISED ME. Totally thought that was all real life footage the whoooole time.
Just kidding.
Using scan data instead of crappy trees and turd clumps that are supposed to be dirt particles made a big difference. These guys probably rock for the geospatial industry. However its been many, many, many years, and they have yet to produce an update that answers a single concern thats been raised. No proof they can animate, no dynamic lighting, no surface definition, etc.
Its incredibly detailed geometry rendered full-bright.
I haven't seen incredibly detailed geometry here at all.
Keep in mind that scan data is usually very very rough. However if you project fullbright photo textures on top of it instead of proper shading and lighting, most of the noise is very hard to notice.
Turning that rough scan geo into something that can work in an interactive environment still requires a huge amount of artist time. I'm not aware of any software that could process the point cloud and reconstruct the lost surface detail, all that can be built from the photo source is already there. Maybe you can smooth it but that would only lose even more detail.
And mind you, that level of noise comes with a LOT of cameras focused on the girl's head. A huge interior would have even less precision.
Covering up this significant issue makes it a scam already. Not having 99% of the features of a functional game engine, starting with lighting or characters, can only cement it even further.
Yea, the problem is it is lacking everything dynamic with the lack of specular direction being the biggest problem. In the forest scene it gets clumsy and again nothing is dynamic which just make it look photo realistic when it is standing still.
I also am very much questioning the claims, that normal games need hundreds of people while they could work with small teams thanks to just scan everything. Yeah that might work for a tech demo. As we see here on PC you can reach a lot with a single artist as long as the scene doesn't ask for anything but look good. And iam doubting, that scanning takes less time, is as versatile and easy to adjust. I mean you have to find places, go there, scan them, work the scan data and make something usefull from them and somehow create an interesting world. I mean the UE4 tech demos had rather small teams as well...full games have a bit more going on, eventhough Warhorse studios show that you can get a lot done in a short amount of time with a smaller team if the software works to your advantage. I feel like this whole "make an interesting world" would be a problem because just recreating real life stuff is not very interesting as long as you are not doing ARMA 4....especially when you can't interact with anything.
I still see Voxel coming to realtime graphics...but not as a full blown engine anytime soon.
For Vanishing of Ethan carter: that is astonishing and looks amazing in-game. Little breaks in flat surface for Stones and bricks really sells the assets. same with uneven surfaces on the floor.
I just stopped the video at certain points of the claimed real life (That looks nothing like real life) and I actually realized how unrealistic it looked like even worse than I thought.
It looks like a environment from someone fairly new to 3d that was dumped into sketchfab or something with little care or effort.
To me it just looks like Google Maps images. They often can look washed out and bland.
I wonder if this process has some of the same tonemapping issues google has with their stitching. I usually see it more of a problem in sunny environments.
Thought I'd add a bit. I used to work for a company that could do the same thing. All this is is spherical scan data with each point getting a pixel from a panoramic camera shot, basically vertex color. Making this into game art is ridiculous, you might as well start from scratch.
There is a difference between what is captured however. What vargatom showed was considered "photogrametry" which is typically way less accurate than a laser scanner that would be used to capture what's in the video. But as far as laser scans go, the information degrades the further away it is, so yeah, this is basically just a pretty picture on shit geometry. Really just 3D reference at best.
Guy kind of comes of as obnoxious with the whole 2 minutes of "I'm wasting time showing real-life footage! Oh wait it's not real-life!"
Does anyone have any idea how much hard drive space something would take up anyways? I feel like storing all of that point data would probably make a pretty damn big file size for large environments, and I don't think anyone would be interested in installing terabytes for a game.
Does anyone have any idea how much hard drive space something would take up anyways? I feel like storing all of that point data would probably make a pretty damn big file size for large environments, and I don't think anyone would be interested in installing terabytes for a game.
Rage was 3 DVDs with heavy compression and got lots of gamer criticism for world texture resolution.
Also, I'm familiar with LIDAR scanning, we've tested it, and I'm not sure if it's even possible to build a complete interior with the tech. You'd have no coverage of the back sides of any objects in the scene, even multiple takes would leave a lot of blind spots.
The way it's used in VFX is to get an accurate measurement of the scene. The crew also takes a LOT of high res photographs from every angle.
Then low poly geometry is manually built based on the pointcloud data, and the photos are projected back on them. Geometry detail is determined by the camera movement, to conserve artist time.
A good example is the scene in Iron Man 3 where Tony is held in the basement and summons pieces of his armor. There are a couple of shots impossible to do in camera, so a virtual set was built based on the real life set for those. They've also used digital doubles and green screen footage for the characters and so on. The illusion never breaks because the camera movement is fixed. Obviously if it was a game scene and you could move around, it'd all fall apart.
Yeah thats basically the jist of it. Unless you can make a game with only fixed camera angles, this sort of scanning tech simply isn't suitable for games.
Its great for architecture work and getting accurate measurements and all that, but not games.
Scanning things is one thing, making it useful is another. Using tons of memory and without nice collisions, physics, animations, dynamic lights and shadows etc. it's useless. These examples looks in it's best like panoramic photo and nothing more. What for is all this detail if it's static. We don't need 3D photography. we need games
Trying to make a game using "unlimited detail" seems a bit crazy and the guy has yet to show anything of this stuff working, especially dynamically. Everythings also very static, and when you get close it seems like the resolution is pretty damn terrible... From afar though, it looks great, and especially statically in some cases it looks real.
Yeah thats basically the jist of it. Unless you can make a game with only fixed camera angles, this sort of scanning tech simply isn't suitable for games.
Its great for architecture work and getting accurate measurements and all that, but not games.
If any of his magic revolves around how the data is stored/manipulated, it could make for a pretty sweet Minecraft clone.
Except Minecraft is all about modifications to the game world, which is kinda impossible if your entire tech is built around a static, highly optimized hierarchical dataset...
There is probably a good reason why Notch called this stuff a fraud.
I think most people that call bullshit on this is not really closed minded just his videos is just too much of an investment pitch, always big words and little facts and also dragging in game developers into the conversation, that after studios used photogrammetry they never wanted to use it again, and that cleaning up the scan took longer time then creating something yourself. So I'm guessing the guy has just given up on looking good to the game industry and is only focusing on capturing the interest of anyone willing to give him money.
I would be more then happy to be proven wrong though
@eld hahaha yeah! It looks pretty clunky to me. The lighting almost looks unlit so it just doesnt compete with current day pbr stuff. That said Id kinda like to see some voxel games again, like the original outcast game, that was wierd but fun.
Their "holodeck" is a just a simple CAVE system, it was state of the art in 1992. Doesn't help that their video keeps on implying the rest of the world is stupid and doesn't understand them.
Well good for them for making the same comparisons to games that ran on last gen consoles 5 years ago. I'm glad they added janky animations, I wonder what they're using to create those, it doesn't look very fluid and no where near what blizzard does.
I don't see why or how the holo-cave requires Euclidion or is made any different by using it. You can do the same thing with polygons and it's not even double input virtual reality.
I still don't like that he bags on everyone that could help him and I find it weird that he complains that those same people aren't helping him. Well yeah, you're kind of a titanic douche to anyone who would give a damn.
I still don't see anything that has the capacity to deliver as rich of a world as what we can currently build. I don't think they'll ever be the revolution they think they will be that completely wipes out polygons but they are more than welcome to keep on trying.
Couldn't they blend this static stuff with necessary animated stuff? "Metro 2033: Unlimited Detail version" "Deus Ex: HR: UD version" "Skyrim: UD version"
Couldn't they blend this static stuff with necessary animated stuff?
Yes, this was mentioned on page 1.
You'd almost certainly get a better looking game if you just used a polygon engine though given that modern engines look better by virtually every measure other than "ok but can u zoom in on a grain of dirt?"
Putting a point cloud system into a traditional game engine would probably be the best approach to accomplish their goals. It's not too far off of Minecraft putting voxel data into a game engine. It seems like the biggest hurdles of sparse voxel octree, my guess at what they are using, is storage and internet speed. Even if they could get a decent sized unique terrain, it's going to be terabytes of data. Maybe they could compress the hell out of the data and get the detail back through some sort of magic procedural generation, but still you don't have lighting, shadows, grass, collision, etc.
Replies
Bruce dell is the ultimate troll, he says these things to ride the wave of angered people. Unfortunately for him the exposure is for naught as the actual product being sold isn't good.
Its incredibly detailed geometry rendered full-bright. Are they claiming this is real-time lighting? From what I can tell they're claiming their models are very detailed, which the are. The rest is capture of a photo. They haven't come out and said specifically thats what they're doing, but I haven't seem them claim the alternative either?
Avoiding to answer direct questions and any criticism due to wanting more investors.
They _are_ claiming that this is actually better than the tech that is currently out there, as they have been claiming for like 10 years.
They are just selling smoke, just smoke... it's their strategic to capt investors, money.
like xvampire said: pure scam, next.
Right, the tech is very cool for certain applications, like engineering, architecture vis, 3d walktroughs (would be cool for the real-estate industry), but it still is a really poor fit for games.
I wasn't that impressed by that video, maybe its the way the guys talking. Makes it sound like a prank tbh.
This video makes it look more impressive
[ame]www.youtube.com/watch?v=DbMpqqCCrFQ[/ame]
Half life 2:
Their claims of "Real Life" (Give some pity because of the compression)
Game from 2004 looks better in my opinion
It looks like a environment from someone fairly new to 3d that was dumped into sketchfab or something with little care or effort.
Just kidding.
Using scan data instead of crappy trees and turd clumps that are supposed to be dirt particles made a big difference. These guys probably rock for the geospatial industry. However its been many, many, many years, and they have yet to produce an update that answers a single concern thats been raised. No proof they can animate, no dynamic lighting, no surface definition, etc.
INFINITE DETAILLLLLLLLL
I haven't seen incredibly detailed geometry here at all.
Keep in mind that scan data is usually very very rough. However if you project fullbright photo textures on top of it instead of proper shading and lighting, most of the noise is very hard to notice.
Turning that rough scan geo into something that can work in an interactive environment still requires a huge amount of artist time. I'm not aware of any software that could process the point cloud and reconstruct the lost surface detail, all that can be built from the photo source is already there. Maybe you can smooth it but that would only lose even more detail.
And mind you, that level of noise comes with a LOT of cameras focused on the girl's head. A huge interior would have even less precision.
Covering up this significant issue makes it a scam already. Not having 99% of the features of a functional game engine, starting with lighting or characters, can only cement it even further.
The scanning is however very cool, the Vanishing of Ethan Carter used it also a lot with great results but of course with traditional meshes and rendering: http://www.theastronauts.com/2014/03/visual-revolution-vanishing-ethan-carter/
I still see Voxel coming to realtime graphics...but not as a full blown engine anytime soon.
For Vanishing of Ethan carter: that is astonishing and looks amazing in-game. Little breaks in flat surface for Stones and bricks really sells the assets. same with uneven surfaces on the floor.
To me it just looks like Google Maps images. They often can look washed out and bland.
I wonder if this process has some of the same tonemapping issues google has with their stitching. I usually see it more of a problem in sunny environments.
There is a difference between what is captured however. What vargatom showed was considered "photogrametry" which is typically way less accurate than a laser scanner that would be used to capture what's in the video. But as far as laser scans go, the information degrades the further away it is, so yeah, this is basically just a pretty picture on shit geometry. Really just 3D reference at best.
Guy kind of comes of as obnoxious with the whole 2 minutes of "I'm wasting time showing real-life footage! Oh wait it's not real-life!"
Does anyone have any idea how much hard drive space something would take up anyways? I feel like storing all of that point data would probably make a pretty damn big file size for large environments, and I don't think anyone would be interested in installing terabytes for a game.
bah
Rage was 3 DVDs with heavy compression and got lots of gamer criticism for world texture resolution.
Also, I'm familiar with LIDAR scanning, we've tested it, and I'm not sure if it's even possible to build a complete interior with the tech. You'd have no coverage of the back sides of any objects in the scene, even multiple takes would leave a lot of blind spots.
The way it's used in VFX is to get an accurate measurement of the scene. The crew also takes a LOT of high res photographs from every angle.
Then low poly geometry is manually built based on the pointcloud data, and the photos are projected back on them. Geometry detail is determined by the camera movement, to conserve artist time.
A good example is the scene in Iron Man 3 where Tony is held in the basement and summons pieces of his armor. There are a couple of shots impossible to do in camera, so a virtual set was built based on the real life set for those. They've also used digital doubles and green screen footage for the characters and so on. The illusion never breaks because the camera movement is fixed. Obviously if it was a game scene and you could move around, it'd all fall apart.
Its great for architecture work and getting accurate measurements and all that, but not games.
Trying to make a game using "unlimited detail" seems a bit crazy and the guy has yet to show anything of this stuff working, especially dynamically. Everythings also very static, and when you get close it seems like the resolution is pretty damn terrible... From afar though, it looks great, and especially statically in some cases it looks real.
If any of his magic revolves around how the data is stored/manipulated, it could make for a pretty sweet Minecraft clone.
There is probably a good reason why Notch called this stuff a fraud.
aah... i wish these guys knew it better... if all it took to make a game was scanning some surfaces with high precision, dang...
Great, more of the same.
answering the easy questions, dodging the technical.
I would be more then happy to be proven wrong though
Pretty much: Old technology barely repackaged. Don't buy into the hype which ought to have died out already 4 years or so ago.
"...Our animation isn't perfect yet but we think its as good as anything you'll find in most blizzard games!..."
Doesn't help that their video keeps on implying the rest of the world is stupid and doesn't understand them.
I don't see why or how the holo-cave requires Euclidion or is made any different by using it. You can do the same thing with polygons and it's not even double input virtual reality.
I still don't like that he bags on everyone that could help him and I find it weird that he complains that those same people aren't helping him. Well yeah, you're kind of a titanic douche to anyone who would give a damn.
I still don't see anything that has the capacity to deliver as rich of a world as what we can currently build. I don't think they'll ever be the revolution they think they will be that completely wipes out polygons but they are more than welcome to keep on trying.
It's not my money they are burning...
"Metro 2033: Unlimited Detail version"
"Deus Ex: HR: UD version"
"Skyrim: UD version"
You'd almost certainly get a better looking game if you just used a polygon engine though given that modern engines look better by virtually every measure other than "ok but can u zoom in on a grain of dirt?"