Wired Magazine: Exclusive Sneak Peek at Unreal Engine 4

[ambershee]

Where DoF really falls down, is it assumes the player is trying to look at whatever the camera is focused on, which is often not the case.

[Baructt]

Where DoF really falls down, is it assumes the player is trying to look at whatever the camera is focused on, which is often not the case.

Agreed. It's trying to achieve an imaginary goal. The only way for DoF to work right is for the computer to have a sensor that follows eye movement to adjust DoF accordingly, which is obviously absurd. If DoF is applied to the reticle, then the odds of it working alright improve. Honestly, though, I'd just rather be able to see everything in full sharpness and detail all the time.

[BmB23]

Indeed that's the big problem. As with anything that predicts what the user wants to do, it *will* fail. The negative effects can be combated by making the effect subtle and predictable. For example if you can reliable assume that whatever is directly under your crosshair is the point of focus, that will be a lot less annoying because it's consistent and predictable.

[Archerx]

The thing is that these effects are supposed to be subtle (HDR, Bloom, Dof, etc) but it seems like a lot of game designers don't understand the meaning of subtle.

[Henrik]

I haven't really followed this thread thoroughly but in film & tv focus is not used to assume anything about where you want to look, it's to make you look at something. DoF is a fine engine feature if used for that purpose, or simulating the limitations of human vision.

[NekoDemon117]

If applied correctly, I can imagine DoF being great to use for when the player is looking through iron sights, reflex/red-dot sights, or scopes on weapons. Then it's obvious that the player is focusing on the sights or scope for better aim. When somebody is looking through a scope, they're not going to be paying attention to the area of their vision around the scope. They're going to be focusing through the scope to line up a shot and snipe out whoever or whatever they're aiming at.

[ambershee]

Henrik - part of the argument for not needing to simulate human vision, is that human vision (with all associated limits) is already being used to view the game. Depth of Field in this instance. In film and cinema, the viewer tends to be more passive, whereas in games it's interactive; in almost all cases, though there are natural exceptions, there is no need to apply depth of field to draw focus to something - especially when the player may decide that that object is indeed not their present focus.

[weephun]

So glad everyone is enjoying the very 1st look at UE4. I can't wait for everyone to see even more soon!

We have confirmed that Epic will release a UE4 UDK down the road, however we do not have a time-frame for that release right now. We're still shipping games using UE3, which is well-supported, and being constantly improved. I love seeing what everyone is doing with the UDK and can't wait for lots more great games/simulators/animations/etc. from everyone here!

[NickP]

@weephun: I can't tell you how excited I am to see UE4 for the first time at E3 this year. Can you or anyone else confirm whether Epic will be presenting independently or not? I'm guessing if they aren't, it may be top secret who's presentation they'll be showing UE4 during. The anticipation is killing me.

[Henrik]

Ambershee the difference is, with our limited human vision we are still just focusing on one thing in actuality: A screen. The simulated environment without DOF would make no differentiation between an object that was far away and one that was supposedly up close. They are both rendered on the screen as sharply and if not for parallax and so on our eyes focus on it all as if it were one object. So what I mean is, a little bit of blur on a distant hill = a good use.. blur on everything besides a foreground character... may or may not be.

[BmB23]

Exactly. It adds depth cues to a 100% flat image. Particularly useful in zoomed views where perspective distortion doesn't do it as much. It helps sell the idea that you're viewing real images captured by a real camera and finally it just looks nice. Win win win.

You're not viewing the image directly, you're viewing it through a terrible display that mangles the image so much. These effects compensate for the display, not for the human eye. The display has limited dynamic range, so you make things overbright (hdr) and make them bloom or lens flare to empasize that thye are bright and not just "white". You add motion blur to add motion cues that smooth out the sense of motion even though the framerate is limited. You add depth of field to add depth to a flat image that itself has no true depth.

[joshua.fontany]

Great to hear that a UE4 toolkit is in the works "at some point".

[Saishy]

I'm with ambershee at the point where you don't need blur because the human eye does it for you.
You can't focus on the entire screen at the same time, which automatically leads to objects not in focus being blurred.

I just hate when games forcefully blur some objects because I'm directly looking at them. You do not move your entire head in real life to focus on a object. Even with the head standing still you can pretty much focus around many different objects with your eyes.

[MonsOlympus]

I find when I look around out in the world I dont see much motion blur or dof using my vision, ofcoarse if youre simulating cinemaphotography things like apature settings and shutter speeds are a common thing. I understand what people are saying about cues but this is one reason I much prefer effects that simulate atomospherics such as image based lighting or height/volume fog. When looking in the distance I dont suddenly lose focus up close unless something is right near my eye, its fairly sharp (up to where my peripheral vision starts) and the distance is hazed by a combination of lights and particles.

In cutscene playback, as these pictures seem to demostrate, the use of these effects does make more sense than in a first person viewport. It does totally depend, artificial lights behave differently to your real light sources such as the sun, my previous responses did cover trying to simulate the eye more closely rather than a camera and they do both react to the differences in light sources and atmospherics in different ways.

[BmB23]

You'll see significant amounts of motion blur if you move your hand at pretty much any speed in front of your eyes while not focusing on your hand. If you had a 1000fps video game this would be true of the game as well, but since what you're really seeing is 30-60 ish perfectly sharp snapshots of the action the low samplerate essentially works as aliasing to remove the effect of blur from your eyes entirely. In fact motion blur can be though of as temporal antialiasing.
DOF is dictated largely by focal length, the human eye is a small thing so you have a very wide depth of field. Cinema cameras that use large film stock and large lenses have a large focal length and as such a very narrow depth of field. If you want a cinematic look you'll want to replicate this. If however you want a cleaner or video-camera look, a wider subtler dof is preferable. But to completely remove it is honestly more of a technical limitation, as is the lack of motion blur. People have just gotten so used to the error that they honestly think it's right.

[Henrik]

I'm with ambershee at the point where you don't need blur because the human eye does it for you.
You can't focus on the entire screen at the same time, which automatically leads to objects not in focus being blurred.

I just hate when games forcefully blur some objects because I'm directly looking at them. You do not move your entire head in real life to focus on a object. Even with the head standing still you can pretty much focus around many different objects with your eyes.

In real life, even a person with the best possible vision can't focus on something too far away no matter if they wish to or not. In the game engine you can when you do look at it, so, some see that as not very convincing. Choosing whether to impose an artificial limitation of any kind in an artificial environment is an artistic decision, and in this case it more accurately simulates real life whether you accept that or not. Whether realism is a pinnacle of gaming is up for debate but most would say that from a psychological point of view at least, the easier you can be tricked into accepting the environment as real subconsciously, the more involved you will be.

[systemic]

So glad everyone is enjoying the very 1st look at UE4. I can't wait for everyone to see even more soon!

We have confirmed that Epic will release a UE4 UDK down the road, however we do not have a time-frame for that release right now. We're still shipping games using UE3, which is well-supported, and being constantly improved. I love seeing what everyone is doing with the UDK and can't wait for lots more great games/simulators/animations/etc. from everyone here!

Woohoo! Best news I've heard all month. Thanks for the update!

[systemic]

Per the DOF and motion blur debate, the key is to use those effects sparingly. When they are done correctly, you (as a player) shouldn't even notice them unless you're a hypersensitive game designer who geeks out on stuff like that. :P

I think of motion blur as an exaggerated way of simulating how your eye refocuses on objects as it moves quickly. Since a player is always looking directly forward (at their screen) their eyes maintain, for the most part, the same direction and so MB does it's best to make the player feel like they are spinning (and does a great job in my opinion).

Depth of field, on the other hand, is the effect that I see overused quite often. There are two ways it should be used - simulating objects that should be in your peripheral vision and simulating objects that are very far away from what the player should be focusing on. The best example I can recall, of proper DOF usage, is when you zoom in with a weapon equipped in the Uncharted series (mainly 2 & 3). When zoomed, the camera moves closer to the player and Drake's head, shoulders, arms and weapon are all *slightly* blurred because they are closer than what the player should be focusing on. More importantly, though, when you're zooming in, your focus goes 'deeper' into the screen, so that moves the character to the periphery of your vision. The second type of DOF should really only be used when you're looking through a lens, such as when looking through a weapon sights or binoculars.

The key, to me, is using is DOF and MB with a subtle hand and letting the lighting and other elements assist in making the vision of environment in your game feel similar to how you naturally perceive things.

[Baructt]

In real life, even a person with the best possible vision can't focus on something too far away no matter if they wish to or not. In the game engine you can when you do look at it, so, some see that as not very convincing. Choosing whether to impose an artificial limitation of any kind in an artificial environment is an artistic decision, and in this case it more accurately simulates real life whether you accept that or not. Whether realism is a pinnacle of gaming is up for debate but most would say that from a psychological point of view at least, the easier you can be tricked into accepting the environment as real subconsciously, the more involved you will be.

Actually, mipmapping achieves the effect of lowering the quality at a distance, thus nullifying that use of DoF. I agree with systemic that there are specific cases where DoF is useful, like mimicking looking through sniper scope lenses or other forms of zoom. Otherwise, I just want my pixels sharp and nice.

[Saishy]

In real life, even a person with the best possible vision can't focus on something too far away no matter if they wish to or not. In the game engine you can when you do look at it, so, some see that as not very convincing. Choosing whether to impose an artificial limitation of any kind in an artificial environment is an artistic decision, and in this case it more accurately simulates real life whether you accept that or not. Whether realism is a pinnacle of gaming is up for debate but most would say that from a psychological point of view at least, the easier you can be tricked into accepting the environment as real subconsciously, the more involved you will be.

When you say something too far away it is really something too far away.
Like a small island on the border of the horizon.

And when that happens even without blur you can't see anything because of the monitor resolution, the pixels will struggle between themselves and blur automatically, thus no need for artificial blur.

I don't know for your but I can see perfectly fine at 1 kilometer, the effect when you can't read or see perfectly small things is achieved automatically in games because of the minimal display possible (resolution).

You can see that when they try to simulate reality there is no artificial blur: http://fc07.deviantart.net/fs24/f/20...y_Timo2141.jpg
And when you do, there is a failed attempt: http://www.swotti.com/tmp/swotti/cac...20cry%2021.jpg

But as they said, blur is fine when you necessarily have to keep control of the user eyes, like in cinematic or maybe a sniper's scope.

[BmB23]

Actually, mipmapping achieves the effect of lowering the quality at a distance, thus nullifying that use of DoF. I agree with systemic that there are specific cases where DoF is useful, like mimicking looking through sniper scope lenses or other forms of zoom. Otherwise, I just want my pixels sharp and nice.

Mip-mapping is a cheap texture antialiasing technique, has nothing to do with DOF.

Game resolutions and particularly game resolutions with a reasonable and playable field of view have nothing on the human eye, or film for that matter. So in a game you can't really see as much as you would in a theater or in real life. Arma notably has a zoom function to counter-act exactly this.

Neither the crysis nor the far cry 2 screenshot has any dof in them, notably because of the scale involved. DOF is a function of the relation of the focal length to the scale of the scene, that is it gets wider the further away your subject is. Which is why closeup macro photography has loads of dof, and microscopes practically have a paper thin slice of field to work with. While lanscape photography rarely has any at all.

I think the portal trailer is a nice use of dof that empasises that the portal gun is super close to the camera while the environment is not.
http://www.youtube.com/watch?v=TluRVBhmf8w

[Baructt]

Mip-mapping is a cheap texture antialiasing technique, has nothing to do with DOF.

Game resolutions and particularly game resolutions with a reasonable and playable field of view have nothing on the human eye, or film for that matter. So in a game you can't really see as much as you would in a theater or in real life. Arma notably has a zoom function to counter-act exactly this.

Neither the crysis nor the far cry 2 screenshot has any dof in them, notably because of the scale involved. DOF is a function of the relation of the focal length to the scale of the scene, that is it gets wider the further away your subject is. Which is why closeup macro photography has loads of dof, and microscopes practically have a paper thin slice of field to work with. While lanscape photography rarely has any at all.

I think the portal trailer is a nice use of dof that empasises that the portal gun is super close to the camera while the environment is not.
http://www.youtube.com/watch?v=TluRVBhmf8w

Just read about mipmapping on Wikipedia. It produces lower resolution versions of textures for displaying at distances. Thus, it lowers the quality at a distance which is similar to the blur achieved by DoF at distance, only better due to a sharper image produced by the mipmapped textures. No, it does not do the same as DoF and was invented for different reasons, but the effect is along the same concept of lowering quality at a distance.

As for Portal, that is a specific case where the effect works for artistic reasons, however I simply don't like the look of the protal gun in that video because I think the DoF is too extreme. I'd rather just see it in normal detail.

[BmB23]

DOF isn't about lowering detail of anything. In fact if you know your sampling theory you'll know you need full resolution source material to make a good looking blur. (and the ever-uglier unreal bloom fails to account for this and produces some pretty terrible blur)

[ambershee]

As BmB23 says, mip-mapping is generally used as a technique for reducing aliasing caused by discrepencies in increasing distance on textures, especially tiling ones (i.e too many pixels in the texture for the given number of pixels on the screen).

[Baructt]

I just ran some errands. While stopped at a traffic light, I closed one eye and compared the sharpness of my stearing wheel while focusing on something in the distance. The amount of blur I saw was so tiny that it was almost nonexistent. I can only guess that DoF used for long distances is supposed to mimick the real world visual defect from stereoscopic vision where we see two objects instead of one due to both eyes looking at something else. If this is the case, then DoF is not achieving its goal since it just makes things blurry instead of creating two. Therefore, it simply mimicks the camera effect and should only be used for such purposes.

(BTW, shouldn't this debate be over by now? Haven't we said enough about it? Yay, UE4!!!)

[Saishy]

(BTW, shouldn't this debate be over by now? Haven't we said enough about it? Yay, UE4!!!)

[Spoof]

Thank you, Saishy. That made my day, and possibly saved my sanity.

[100GPing100]

I just ran some errands. While stopped at a traffic light, I closed one eye and compared the sharpness of my stearing wheel while focusing on something in the distance. The amount of blur I saw was so tiny that it was almost nonexistent. I can only guess that DoF used for long distances is supposed to mimick the real world visual defect from stereoscopic vision where we see two objects instead of one due to both eyes looking at something else. If this is the case, then DoF is not achieving its goal since it just makes things blurry instead of creating two. Therefore, it simply mimicks the camera effect and should only be used for such purposes.

(BTW, shouldn't this debate be over by now? Haven't we said enough about it? Yay, UE4!!!)

Blur isn't just the creation of 2 images. I can see blur in objects that are only visible to one eye.

[BmB23]

I just ran some errands. While stopped at a traffic light, I closed one eye and compared the sharpness of my stearing wheel while focusing on something in the distance. The amount of blur I saw was so tiny that it was almost nonexistent. I can only guess that DoF used for long distances is supposed to mimick the real world visual defect from stereoscopic vision where we see two objects instead of one due to both eyes looking at something else. If this is the case, then DoF is not achieving its goal since it just makes things blurry instead of creating two. Therefore, it simply mimicks the camera effect and should only be used for such purposes.

(BTW, shouldn't this debate be over by now? Haven't we said enough about it? Yay, UE4!!!)

As I said the human eye is a small thing, and as such has a very short focal lenght, and a very wide depth of field. It is mostly noticable with extremely close objects, such as putting your finger up in front of your eye all the way.

As BmB23 says, mip-mapping is generally used as a technique for reducing aliasing caused by discrepencies in increasing distance on textures, especially tiling ones (i.e too many pixels in the texture for the given number of pixels on the screen).

To be theoretical, it downscales the texture, which is the same as running a high pass filter/blurring/antialiasing. And it does this as a preprocess step which makes it much cheaper than taking extra samples. In fact it'd be ridiculously expensive to take these samples at runtime. see: supersampling. Of course since polygons take a new transformation every frame you can't precompute that (although temporal aa seeks to exploit that they dont change too much yes), and runtime it's only linearly interpolated which makes it blurry at high angles, hence anisotropic filtering is used that only takes new samples along the deformed axis.

[NickP]

Can we please drop the depth of field debate and get back on track? This was meant to be more of a speculative thread regarding the Wired article specifically, as well as the E3 announcement that will follow. The technical discussion is getting a little out of hand.

With that said, let's get some relevant discussion going. I'm really interested to know how you guys feel about Unreal Engine 4 debuting at E3. Do you think this means the "debut" will be an actual game instead of a tech demo? Will they debut during one of the console maker's presentations? If so, who will it be and why? I'll get the discussion started with a few thoughts of my own.

Back in April we heard news that Epic Games was working on a PC exclusive title, much to my salivation (and others, I'm sure). I have to wonder if this title will be the focus of their E3 announcement? If so, I'm sure we're all thinking "Unreal Tournament 4, please!" That would be exciting to say the least, but I have another more far-fetched idea that has me thinking...

In every interview I've ever read about Unreal Engine 4, Tim Sweeney and others specifically call out Sony and Microsoft -- but never Nintendo. It's always "Sony and Microsoft have to step up their game to make this work." To me, this could mean one of two things:

1. Nintendo is not even in the picture.
2. Nintendo is already on board.

With the rampant rumors of Wii U running Unreal Engine 4 and Epic remaining quiet about where and when at E3 we'll see Unreal Engine 4 revealed, it's pretty exciting to think that they might do it during Nintendo's presentation on June 5. I have to admit, if this were true, I'd jump for joy at the possibilities of such a partnership. But enough about my crazy theories. What are your thoughts, everyone?

[BmB23]

To be honest the Wii U seems like kind of a bad idea and I'd be sad if that was the only console platform that could run it well. I'd much rather just have a much more precise wiimote since it never really worked right.

Anyway the blur discussion sprung from the fact that Unreal blur kinda sucks right now and I'd like to see that improved in the future. Then someone has to say lol dont need blur.

[NickP]

To be honest the Wii U seems like kind of a bad idea and I'd be sad if that was the only console platform that could run it well. I'd much rather just have a much more precise wiimote since it never really worked right.

Anyway the blur discussion sprung from the fact that Unreal blur kinda sucks right now and I'd like to see that improved in the future. Then someone has to say lol dont need blur.

I personally think the Wii U looks like a great console and is oozing with potential. Let's just hope Nintendo announces a strong online component, launch titles, and great pricing. Who knows what Sony and Microsoft have in store for their next consoles. We may not find out for a while.

I do understand where the blur discussion started; I just wanted to put an end to it because this wasn't intended to be a technical discussion.

[Baructt]

To be honest the Wii U seems like kind of a bad idea and I'd be sad if that was the only console platform that could run it well. I'd much rather just have a much more precise wiimote since it never really worked right.

Anyway the blur discussion sprung from the fact that Unreal blur kinda sucks right now and I'd like to see that improved in the future. Then someone has to say lol dont need blur.

Actually, the Wii Plus remote works great, so Nintendo already fixed that. Zelda Skyward Sword demonstrates that quite nicely. The Wii's controls can work nicely for shooting games, too, if devs would just design a game to use them right. Using a bounding box where the reticle moves around freely while moving the screen at varried rates would be a great system for most Wii shooters, but few actually use it. As for the Wii U being a bad idea, I simply can't comment until I've seen it in action more. I do think that Nintendo badly needs a standard controller for their console instead of motion controls. They do have the classic controller, but I question how many devs would use it for new games out of fear that their audience does not own one. Nintendo has been getting things wrong ever since the N64 came out. What are the odds that they'll do it right this time?

I'm excited that Epic is working on a PC title. I hope that it's not another gory game, but that's kinda their specialty. I might buy it anyway just to see the best in PC graphics (maybe).

[RNG]

I think the PC exclusive title they're working on is a F2P game. I'm guessing they want to jump on that bandwagon before its too late. Their trailer for Fortnite looks like its extremely susceptible for F2P model in my opinion. Since Fortnite is an UE3 game, I don't think we'll see more of it in E3, even if we do see it, it won't be relevant to UE4 obviously.

About Epic pushing the console manufacturers, well, one has to remember, during the previous generation, xbox didn't really had any significant market share. Playstation 2 dominated the entire market. While launching XBox 360, Microsoft wanted it to be a competitor to Playstation 3 this time, thats why they rushed it (with problems such as rrod) and actively listened to the developers even though it costed them billions of dollars. This was before recession as well.

Fast forward to now, giant recession, XBox 360 has a significant market share, Playstation 3 is behind it, but still close, and Nintendo has the whopping lion's share when it comes to the install base in this generation. So basically Microsoft and Sony sold expensive top of the line hardware (during its launch) at a net loss per unit, and neither was able to grab the top spot, while Nintendo, with its gimmicky control and cheap hardware, became the market leader and made a net profit on all the consoles they sold.

To be honest, market has changed a lot. I don't think Epic will be able to convince manufacturers to increase their specs. Neither Microsoft nor Sony wants to sell new consoles with a net loss. Sony can't really afford it, Microsoft can, but it doesn't mean that they should. At best, maybe Epic can persuade them to increase the memory size, but the rest of the specs are probably set in stone at this point.

Nintendo never really had great relationships with 3rd party developers. They were always pain in the ass to work with. So I wouldn't be surprised if they aren't even actively talking with developers.

[darthviper107]

I would think with the controller being a high cost that Nintendo can't afford the graphics power that would be needed for UE4. At best I think they will be able to run current-gen UE3 games at 1080p.

And from rumors, it looks like the graphics capability is just a bit above the PS3 and Xbox 360.

[Baructt]

I would think with the controller being a high cost that Nintendo can't afford the graphics power that would be needed for UE4. At best I think they will be able to run current-gen UE3 games at 1080p.

And from rumors, it looks like the graphics capability is just a bit above the PS3 and Xbox 360.

Nintendo has always simply said that the Wii U would be more powerful than current gen consoles, as if suggesting that it was made to be a competitor with them. Unfortunately, that doesn't say much about how it will compete with the next gen.

@ RNG, Nintendo has always tried to get 3rd parties to go with their console and has always failed. For that reason, they will always keep trying to get them on board. That's why they made the Wii U a current gen console instead of worn out tech. As for the past, Nintendo has always made mistakes with their consoles that drove devs away from it. The Wii is under powered, the GameCube had small discs, and the N64 used expensive and awkward cartridges instead of CDs. I'm starting to think that Nintendo's only real strategy is to be different from everyone else.

[RNG]

For those who wish to see the GPU particles in action: Linky

Thats APEX Turbulence running on UE3. Its probably similar to what they're running on UE4.

Video description quote:

APEX Turbulence in UE3 test level. The energy particle effect (blue) is using 36K particles, the character particle effect (green) 41K particles and the grenade (orange) 100K particles. So a total of 180K particles running in UE3 on a single GTX580 above 60fps.

Actually, it seems this is the exact same thing what they'll use on UE4. Starting at 2:00.

[BmB23]

Protip, you can link a specific time by appending #t=5s or =2m30s or 5h3m15s etc

for example http://www.youtube.com/watch?v=ul3v2K3cXK4#t=2m

[RNG]

Protip, you can link a specific time by appending #t=5s or =2m30s or 5h3m15s etc

for example http://www.youtube.com/watch?v=ul3v2K3cXK4#t=2m

You can also do that by right clicking and selecting copy url with current time, thats not the point.

[MonsOlympus]

You can copy/paste the link from a specific time too, always wondered how people missed that feature.

I'm not sure about this whole nintendo thing, Epic's always been very vague when it comes to supporting them. They might show something on a wii U but it could be a game product on UE3 as they have a number of teams under their roof.

In terms of the content we are seeing, I remember seeing GoW content in the early UE3 demos but this demon knight guy looks like some high res infinity blade character but the environment art seems new so perhaps there's something in the works that's new IP. I remember Mr B having abit of a speil bout how new IP is great.