Art Direct for 100 MPH: Balancing Visuals with Performance for Forza Motorsport

[Music] thanks everybody for joining me today my name is Matt Collins and I'm the CG supervisor for turn 10 this talk today is art directing for 100 miles per hour and this is really talking about the unique challenges that we have art directing for a title that moves so quickly we don't have a lot of performance bandwidth because we have to move at a hundred sometimes even 200 miles an hour so we have to be really focused on sort of the core art values and techniques for for how we manage our frame so a lot of the talk today is actually speaking to the core art principles and core values of art but because we don't have a lot of sort of bandwidth for fancy graphics effects and everything else we really have to double down on art direction and those core techniques so I'm gonna go ahead and start going through it the first thing

we're going to show you is a video a trailer for our game it might be a little choppy all the videos that we played today might be a little choppy so just enjoy their visual excellence but know that they might stutter a little bit so we'll go ahead and fire that up [Music] [Music]

[Music] [Applause] [Music] [Applause] [Music] [Applause] [Music] [Music] [Applause] [Music] that is our very pretty game which I am very proud of it came out a couple of years ago and we're doing even even greater things now so it's been a lot of

fun to work on so the overview today is going to be we're gonna do a brief history of Forza and talk about the unique challenges that we have making a racing game and how we prioritize visual investments for those challenges that'll be the majority of the talk I have a few slides in here this is originally a GDC talk that talks about how we manage our performance budgets and how we manage our our content budgets and that might be less interesting to you guys I'm not sure so I'll go through it relatively quickly but if there's things you want to ask me about or dive deeper on feel free to ask questions later in the talk and then I'll talk about some of the pitfalls and key learnings because there's always key learnings after you develop a title that we've had and then summarize the talk at the end so let's jump right into it yeah so my name is Matt Collins I spent four years at from the state universities University studying illustration I actually wanted to be a comic book artist and right as I graduated the industry the comic book industry fell apart it's I had to

scramble and figure out what I was going to do after that so I wound up going to DigiPen I spent two years at DigiPen and they gave me a very focused career or focused education around games and then I wound up as an intern at Nintendo and eventually getting hired there full time I was at Nintendo for three years working on waverace 1080 puzzle league and eventually moved on to Microsoft where I've been making forts of titles since four to one and I've been there ever since and I'm now the CG supervisor at turn 10 so yeah just talking a little bit about the history of Forza we've made seven titles in our main motorsport franchise there's also four titles in our horizon franchise which is part of our sister studio playground they're over in the UK so we've also worked with them to make a number of those titles and we also have a mobile title for its so a lot of years of forza

so talking a bit about the challenges that we have we run at 60 frames per second which in itself is always a challenge it just doesn't give you a lot of margin for error when you're rendering our cars move it over 200 miles per hour at some in some cases specifically the Bugatti Veyron which has been the bane of my existence since that guard came out we have photo realistic visuals which is a difficult target to hit it's also kind of a plus which I'll talk about in a minute we have what long draw distances so in some cases we can see up to three miles from an area on the track from one end to the other so that's difficult to manage and most of our tracks are covered in vegetation so this makes it difficult that's hard vegetation is notoriously hard in performance so that's the challenge that we often face and also in terms of composition our options are limited just because we are using real-world environments so next I wanted to show you a video that sort of illustrates the main difference between our game and say something like a

first-person shooter like Halo so what you can see here is if you look at the scene on the right that's our game and the bugatti veyron moving through our longest straight in the game on luma and you can just see how quickly you move through the scene so this puts a lot of pressure on the system we're on the left if you have something like halo where you know it's still obviously there's a lot of graphical beauty here but you're moving through the scene at a different pace which just creates different bottlenecks on the system so often times when we're moving through the scene like this we'll have problems where the textures drop out the Geo drops out the whole world drops out the system just has a hard time keeping up with how quickly and how rapidly we're throwing geometry and textures at it so it's always been a challenge for us and so we what that leaves us is not a lot of performance and I'll talk about that so the perks of making a racing game it's not all bad it's a controlled experience it's on Rails we generally

know which direction that you're moving in we can run some of our off-track modes at 30 miles per hour at 30 frames per second and the fact that our cars move over at over 200 miles an hour is kind of a boon at times they means that they can't the user can't slow down and and look at every minor detail and it means that we can spend more time on the overall frame and less on the individual quality of an asset because we know that person is moving very quickly through this to the seed and also photorealism the nice thing about it is that it's a clear style choice we know we're trying to hit it's not ambiguous a lot of people have a frame of reference and when we're outsourcing so that's helpful to us so how do we prioritize visual investments at speed and like I said this goes back to a lot of the core art principles I like this quote here the first principle of the art is not to rely on tricks of technique most swordsman make too much of technique sometimes making it their chief concern what we've seen is when you have the

ability to use a lot of fancy graphics effects it allows you to sort of get more relaxed about some of the core art principles art values because we can't rely on those effects we almost are exclusively on core our principles and values and so this provides us a razor focus and we actually believe that these because they're so core to art that any game title or anyone that's making something if they pay attention to these core principles it just makes their work better and oftentimes we forget that we see that with students we see that with professionals that the computer gets the better of them and they rely too much on sort of the whiz-bang stuff and not enough on those core fundamentals so when I talk about the core fundamentals which is mostly what I'll be talking about today this is what I mean talking about composition value palate texture shape storytelling

cadence and performance and these are the topics that I'll be going over in further detail today and something that you'll notice is what I tend to do is move from large to small this is how you should build up anything like a painting or at any art piece really a lot of people will tend to get focused on texture too quickly without having settled these other core principles and so as I go through this talk I'll kind of be moving through it from that big suit with big to small in mind so starting will talk about focusing on what matters we spent a lot of time thinking about composition and I flow just because we generally again you're on Rails we generally know where where you're looking at what you're looking at however early on in our titles we weren't very good at this because a lot of us were working used to working on first-person shooters or other titles and we focused the majority of our attention around this area so we've put a lot of detail into our trackside objects into our into our buildings and

what we found was it just didn't have a lot of bang for the buck that just doesn't take much of the real estate what we call it is the wedge that this takes about 80% of the real estate of the screen so it's critical that the read of this is the most important part of the whole scene the sky the road getting that right that's everything and you can let a lot of the stuff go because we know that this is taking up so much real estate in our scene and is really what the player is looking at we spent so much time thinking about this that what we did was we developed a system where if you're familiar with the rule of thirds where compositionally you split your frame into three parts we did that here and then we mapped out where the intersection points of those thirds were and then we took it a step further and what we found was that the focus of the viewer the focus of the player tended to be in this sort of red diamond in the center and we even backed this up with we had connect eye tracking at the

time and we looked at a whole whole group of players to see where were they generally looking and it tended to fall right in this area so we know that when you're driving this is the majority of where your focus is so this caused us to prioritize our scene accordingly our p1 all the assets where we would spend the most amount of time the most amount of Polish bug-fix would all happen within this area this would be our p2 area the Sabir at the area that we would fix things but it would be less critical and what we found with the p3 area was you could almost ignore it I was amazed at the bugs that we would have in this area that seemed glaring to me because I was used to staring at the screen that people wouldn't see at all whole objects dropping out the world missing it's just when you are focused on an area it's very difficult to look anywhere else and people are just not looking out there so it really just shows you how much composition matters and really not spreading your efforts out like peanut butter so we use this tool so much that

we actually created an in-game overlay that allowed you to see where you were looking and we would have this on screen at all times so that as we're driving around we can really see where is the focus where is the player looking and where do we need to focus on polishing our assets the other thing that we noticed is that this focus did tend to shift depending on the context of the gameplay so if you look at this area here this is a user that's taking a difficult turn in the wet and so as they're turning their focus changes from being out in front of them to right at this corner you naturally start to look down at the at the corner that you're taking your focus and your attention goes to that because you're in a high intense gameplay moment this here is where you're straightening out and what we found is then your focus shifts it goes from being at the corner to up and out towards the horizon so I'll talk about this later but knowing that we tended to put our big moments or our big areas of visual interests in this case having a really compelling sky out in front of you because we know

that's what you're looking at at that time so talking about the corner what we did is if we know that you're looking at the corner at a particular time that's where we'd spend a lot of our attention and focus in detail down on that corner so then we would start to add things like storytelling elements extra detail we'd spend a lot of time and attention on that corner because we know that's where your attention is and knowing that really helps us prioritize as we're directing our content throughout because we just don't have a lot of time to make content at the same level everywhere so we really have to understand where the player is looking and focus our attention there we like to think of this focus like a painting and how you would approach a painting if you look at this which I really love this is a painting of a woman that from by one of my favorite artists where all of the attention to detail is right around the face where your eye tends to focus anyway the core principles of value and saturation and all that and palette is

still really good around the scene but there's nowhere near as much detail which you might not even notice at first as soon as you get away from the face the detail just drops away because the artist wants you to focus here and what we found is if you actually put detail everywhere it confuses the player then are the keys of the viewer then they don't know where to look so composition and attention to detail really matters and especially not putting the detail everywhere in the painting is really critical and for us especially because we're also limited in time and the amount we can spend on each asset so next I want to talk about the frame and how we designed for the frame because we're moving so quickly your your the frame is really your impression of the scene this is the part that we have to get photorealistic and one of the biggest tools that we have to getting photorealistic visuals is around value this is also true if you looked at at paintings as well so much so to the point that we would actually do blurred grayscale thumbnails of our

scene just to make sure that our values were reading okay and that nothing was standing out to actually call out this scene in particular we used to really struggle with this where we would have areas like our guardrail if I show you over here or areas like our Road lines being too white or not not cohesive and what that does is contrast draws your eye and what we found is that people are actually looking at the road lines looking at the guardrails rather than looking at the car and this created a non cohesive scene and it created sort of a discordant swith with the viewer and so we really want to make sure that we're balancing our values make sure that they land in a photorealistic frame how we do that if anyone isn't familiar with this is with the concept of authoring our textures color correct so what we use we have the signal the color checker here these are industry standard values and what we'll do is we'll take a picture of something for a texture reference and then we'll bring it into

Photoshop what happens when you take a picture of something in the real world is that it's colored by light you can see that these are two very different scenes one is in the overcast as one is on a bright sunny day and what we're trying to do is we're trying to remove all of that light information all of that color information and get down to what is the base value for that color and so we'll bring our textures into Photoshop and balance them because Photoshop knows what color this this color checker needs to be and they'll balance it to get the appropriate values and this is key for us to get good values and good base values at the beginning when we're authoring our textures books so we even created the spectrum that shows these are where all the things in life this isn't all of them but is a handful of common ones that we use in our game fall on the value spectrum and what this shows us is a couple of things one is there tends to be what we call the art direction range almost nothing in real life goes above doesn't go to pure white and never goes to pure black

if something goes to pure black it's like some weird special material that's designed to absorb all light but that rarely ever happens in real life so we'd never allow things to go over these values the other thing we found if you look see this red line here is that most values fall on the darker end of the spectrum and we actually see this error in in when people paint as well they tend to author their their values or their diffuse colors to light because they're used to seeing them hit by light but what happens is you want your diffuse color to be dark to allow your your light to brighten them up later and so this is a common mistake that a lot of artists make is to author to write because they're used to seeing values in the Sun so this is something that we control very heavily this is to show that same image in grayscale to also show how much color can fool your eyes if you look at those red dots they look like they should be much brighter because you read red as bright but their

value is actually very dark and so we do this because we also know that you can't trust your eyes when it comes to value or when it comes to sunlight so we spend a lot of time with this I spend a lot of time talking about it here because honestly it is one of the most critical elements of getting a photorealistic scene so we spend a lot of time working on it this shows the value of that this is an in-game representation this is an early shot of our scene ignore there's some broken light maps on the road but if you look at the building it's showing you that almost none of the values are correct here the lighting isn't done right we don't have any bounced light we don't have good value that a fuse colors aren't correct and the net result of that is this doesn't look very believable it looks very CG it doesn't look right to your eyes once we've done our pass when we get the right color balance the right color correct we get the right lighting on the scene we get the right bounce this is these are all the things both in our game and when you're painting or doing anything that

add up to being a beautifully lit scene that looks photorealistic to you and so we spend a lot of time working on this and this is a really good representation of how much getting those values right matters just another example we so if you look at the images the two images on the top our photo the one on the left is actually our old studio we did this because we really want to spend a lot of time going out and rechecking our values the image on the bottom is from our game and you can notice a few things that are slightly off like the trees aren't great and maybe they're reflections but you can see how closely and how much time we spent making sure that our values we're as close as possible and then that translates the two images on the on the right that's Laguna Seca the image on the top is a photo and the image on the bottom is our game and we just spend a lot of time again really massaging those values and making sure that we get them right to produce a photorealistic scene and this is just an example of that same

scene in Laguna just with its value so you can kind of see where we landed with that so the next thing I want to talk about is palette going kind of going along and that we move in a very sequential manner when it comes to palette palette was what we use to create our mood so just a quick anecdote here we so we did this we shot Dubai in the desert when it broke the heat record for that area for all time so while we were out shooting in the Sun it got up to a hundred and thirty-one degrees so all of our we were out there all of our cameras stopped working right around one o'clock in the afternoon everything just just shut down it was just too hot including us so we went back to our car and what we found in our car was that everything that was metal or thin had actually melted so all the little metal springs all the coils all the little metal bits in our camera cases had all melted into these little puddles of liquid metal so

we went home because we're like alright that's that's probably too much and we actually pulled the key cards out of our pockets which were plastic and they had all melted too so we kind of threw them on the desk of the hotel and we're like wow how do you guys deal with this and the hotel owner he's just like because no one goes out in the day at this time are you guys idiots so we really really struggled here but what we really found was there was a mood and atmosphere to this that we wanted to capture the sense of heat the sense of intensity and the way we wanted to capture that is through this palette what we really liked was this feeling of this warm orange sand against this supercool blue sky we felt that it was a really great juxtaposition and a great way to show the intensity and the heat you can kind of almost feel it right and so you can see at the bottom we also have what we call a weighted palette which is to say the larger blocks are the ones that are going to take up more of the scene and they're gonna have a greater effect on the scene and so this is something we

really think about what we're trying to establish mood this is another shot of a real-world track and us basically doing the same thing but really thinking about palette in advance I'll show you some examples of what the impact that is this is one of our tracks catalunya and what it is this is it just vanilla right out of the box and what we want to do is we want catalunya have more of a mood we weren't sure how and we actually used palette and shape language which I'll talk about in a minute to establish that so we talked about what if this had like a real high intensity f1 racetrack mood and we did this so through shaped language and through color this is very stark almost black and white in terms of its you know it's it's very red and white almost monochrome but it feels very professional very very race like and then we said well what if this was a bit more festive and we looked at this and so you can see and this also if you look at shape language as well you can see between powell in a shape language if you look again how much this palette

is having an effect on the mood and the complexion of the scene and so these are the types of things that we spend a lot of time thinking about what mood do we want to impart to the player what impression do we want you to take away and that really shows you how much palette has an effect on your impression of that seed I talked about shape language but that's another thing that we think a lot about the shape language really helps determine also the mood and the feel in this case what we liked was that Dubai is actually located right on the ocean which not a lot of people know and there's a lot of sailboats and things like that we actually thought that motif would be a really cool juxtaposition against sort of the harsh dry rock-like atmosphere of the desert so we created all these really nice sail shapes these are concepts and we applied them to all of the objects in the scene so we carried this through our tents through our grandstands this idea of shapes sails very tent-like structures and here's an example of those structures juxta bucks to pose these are is a concept against the scene and you

can see how much that shape language impacts and plays against excuse me against how we how we compose so really great values there and this is a final concept that we did that kind of pulls all those elements together here's a palette that we use for the scene here's you know kind of pulling that shapes into the into the buildings into everything just to give you that that overall feel so it's something that we think a lot about when we're making our seeds all these things is to say that we're trying to impart a mood and atmosphere what's really interesting to me is when this is Prague this is one of our crown jewel tracts when we went here we actually were really at first really attracted to this one image that we had that showed Prague it was almost grayscale it showed that Aston Martin it looked really haunting but when we got to Prague this was actually the atmosphere that stood out to us this kind of golden light these Renaissance colors this mood was with us the entire time we're here it had this really nice

haze and we're like oh this is the mood we want to capture we just love this feel and this is often why we send our track leads on reference trips so they can take away this sort of feel the smell the mood that you really don't get from looking at pictures a lot of the time and so we make sure that we impart that mood into all of our tracks another little interesting anecdote here this is the charles bridge so we took a laser scanner out onto the charles bridge and if you if you know what a laser scanner is at the time it was six feet tall it looks like a giant evil robot and we put it we set it up at about 4 a.m. in the middle of this iconic bridge structure and i don't know if you've ever set a large piece of technology in the middle of an iconic bridge structure at 4 a.m. but it tends to freak everybody out so while we were in the middle of laser scanning even though we had permits cops started running at us from both sides of the bridge and so we were frantically trying to get the the scan to complete before the cops caught to us and we just we got

the last rotation right before they got to us and we said oh we're sorry here are permits we apologize we picked up everything we moved on but yeah this was a this is a real challenge to shoot so anyway referencing for tracks is always entertaining but these are just so moving on these are just other examples of mood and atmosphere for a track this here is showing daytona this is also a very different mood hot white heat packed people this is high noon very intense shadows so at a very different mood this is the Nurburgring so sort of haunted grayscale just very very different mood in itself and so these are the things things we think about when we go to more tracks I'm just showing and this is concepts that we do afterwards to translate those who didn't go into these mood boards to give the player a feel for all the things we talked about haze color palette light temperature all that and that's what we use to create our scenes we use things like fog here and

sky this guy has a huge impact on everything in your scene you've probably seen that when you paint when you do anything so really paying attention to the sky and what kind of mood you want to part with that is a big deal for us so I'm moving on to painting with light I'm gonna take a quick drink though this is another thing that we do quite a bit of is we use light not often the light scenario that we have is not meeting our needs or it's not giving us the effect that we want so we spend some time changing our light I'm using our light in this case this is an early concept that we did the problem we were having with this tunnel was it was too dark and it was really difficult for you to get a sense of parallax because there was no contrast for you to be moving again so it felt like you weren't moving very fast so what we did was what we often tend to do in these situations is we actually started to cut holes in the roof we started to move the Sun direction to get more contrast on the

scene to get more visual interest to get that sense of parallax this is us continuing to experiment with that we still weren't getting what we wanted until we finally landed here this gave us the right amount of light it gave us the right amount of visual interest the right amount of contrast and gave you that sense of speed so we'll often do this we'll either add or remove objects will add tree wells will do things to use light to our advantage to give the effect that we want to give moving on I'll just touch on this real quickly one of the other issues is with scale what I don't actually have here which I apologize for is an example of bad scale but this is an example of good scale what we found early on when we had so few Polly's to deal with is that we would spend all of our Polly's on our big objects and ignore our little ones and what happened was that actually made everything in our scene feel really cartoony and you also didn't get a sense of speed when you don't see these small grounded objects that the player is familiar with then there's no sense of context there's no feeling for how big a thing is what its scale is or how

quickly you're moving against it so we actually had to double down on all these small grounding elements all these small set pieces so if we have a building including the windowsill if we had a race I'd object including these like kind of these booths and these carts because we want to give you a sense of context and why and how big something is and so we actually double down on these we actually spend less Polly's in our large building and our large assets and more Polly's on these small grounding elements to give you that sense of scale and proportion this is something that we see often that beginning students don't do well they forget to add these grounding elements and as a result their scenes feel cartoony because you don't know how large their so finally I want to talk about texture and again I saved this one for last because while this does add a great deal to your scene it tends to be the the area that students and even professionals tend to jump to too quickly and so texture is obviously important to give you that sense of

detail and surface type what is this thing going to feel like when I touch it but you don't want to go here too early so if I don't even if anyone is familiar with photogrammetry this has been a groundbreaking technique for us and something we leverage heavily now but this is the where you take a series of images with your camera and then you use a program like a geosoft or reality capture to assemble those pictures into a 3d model and so this is an example of a model that we did here this is a tree using photogrammetry and our artists build almost all their assets with with this technique at this point and it's been really awesome for us this is another example of where we created this rumble strip using photogrammetry and it just gives you and see all these storytelling elements all this all this really nice texture this is these are all the little elements that add up to making something look photorealistic in this case we are using a photo but is often the details that people miss when

they're creating those textures is all those little details and story elements so now I'm going to show you all this in practice these are a few before and afters this is obviously before where we forget this all the time as well and so oftentimes what we'll say is you know what the scene is not working we're not using all the principles principles that we just talked about and we have our concept artists actually go in and do paint overs to help guide us this scene right here is doing everything wrong we're lacking those elements of scale it's all one value we're not painting with light we're saturation isn't good it's doing everything that we we just talked it's not doing everything that we just talked about so we say help us out concept and they'll create this image and you can see what a difference this makes in terms of value in terms of visual storytelling in terms of painting with light it's doing all those things I talked about and the difference is stark and we have to do this all the time because we forget as well and it's something I often recommend

students do is when they get about 3/4 of the way through the piece take it into Photoshop paint over it's really easy to iterate in Photoshop and we find that it helps break that sort of disconnect that you have in CG and so we'll do this a lot this is another example barring the terrible crowds that are over the bottom right again scene feels it feels okay but it feels flat it doesn't feel dynamic it doesn't have those elements of storytelling you can see the reference image down in the left and even the thumbnail read is better than what we have so we took it in and we said what can you do concept and they produce this image again it brings that those about that value separation it brings that contrast they even put in smoke and birds and all these little details that we pre forgot about so highly recommend doing this and it really helps to illustrate all these points that have been talking about so yeah just talking about the cohesive frame this is none of this is new these are all core values that everyone knows about in in art but we tend to forget but is a combination of all these things

line shape color texture value form and space and it's easy to forget them but it's they're just as important in CG as they are when you're doing a painting so the next thing I want to talk about is cadence how you move through the frame so everything I've talked about now is still frames but we also want to talk about how we design the experience when you're racing through a track early on we did this poorly when we designed our race tracks this is a concept that we did before we did one of our race tracks called Rio and we said is we're gonna put a visual moment something to blow you away on every single corner of the track and the problem is is everyone knows when you turn everything up to 11 nothing is and so what we found was that the player had no areas of visual rest they were overwhelmed and they weren't getting the impact of a lot of the areas that we designed so this was an important thing for us this is a track that we designed later this was Dubai

and what we said was we're gonna have four key moments four areas that we do turn up to 11 and allow mini moments or areas of rest in between so we like to think about this as a piece of music you want high points you want mid points you want areas of rest it's often how you look at a painting you want to make sure that not everything is up to 11 and so now we design our tracks and in my opinion they're much more successful because the key moments that we design have much more impact to the player we also don't spread our detail over the whole scene like peanut butter this is a mistake that we made early on what we do is this is a shot from our game this is the iconic hotel from yes Marina what we do is we design these areas to have the most visual impact and we actually put about 80% of our art efforts on these areas because we know as you're driving around the track there's the hotel again that that is what you're gonna see the majority of the time and that's what's

going to have your visual impression and we left the other stuff just kind of bleed off into the into the distance this is an example from Rio this is basically what I was talking about before where if we know that this is on at the end of a long straightaway we built this iconic moment this a shot of Sugarloaf Mountain and we know that we have your attention here this is a low impact game player you're going straight you're going downhill there's not a lot that you have to do here and so we often will design our racetracks to build in these straights so that we can focus your attention on these moments so you can appreciate them so you have the time to appreciate them and then when you take these high-intensity corners you know we we sit we we just spend less time on them we put our details in around the corners instead but really designing this experience knowing what your player is going to be doing and putting your detail as appropriate and this is just again talking to we also do this with our gameplay we actually design skids impact moments things like that because we know where

we're gonna have your attention and where we can spend our attention to detail this is an iconic turn at the end of Laguna Seca and so this is a turn that everybody misses and we wanted to tell you that story and kind of forecast it for gameplay to say this is difficult turn and you're probably going to miss it too so we do our skids we do this sort of like dirt run off to say a lot of action happens here a lot of people take this turn run and we want to forecast that and so we tell you this story when we're deciding our tracks so again these are all different examples where we pull some of that visual storytelling into our tracks when we're designing that cadence and it just pays off for us and in a big way and again like I mentioned we designed it like a piece of music this is a waveform of Beethoven's fifth so it's just really interesting it shows you throughout the piece they have their highs their lows they kind of have two big centerpieces and then they have areas of rest and

this is how we wouldn't want to design our experience as you move through it so the last thing like I mentioned we're getting low on time here so I'll move through this pretty quickly there's a lot of charts in here so try not to glaze over I'm gonna move through them pretty quickly but what this piece is really telling is about we have to spend a lot of time thinking about performance so we're not just designing our tracks and designing our visuals in a vacuum we have to think about GPU and CPU and all the limits that we have but we actually find it sort of enables us in a way this is the sandbox that we have to play in and we have to be creative within it so these are just for anyone that's interested these are the budgets that we have this is only to show this 2.8 number is how much GPU for we have for our tracks for anyone who's familiar with GPU budgets that's not a lot of time so that's just highlighting how little time you have with it these are these are tools that our artists used to

evaluate budgets this is just a chart showing that anywhere that's over performance turns red and our artists have a nice visualization of the track to say those are the areas that we have to work on and this is what we're gonna bang at to to improve our performance and this is a great tool for us so how do we prioritize there are these visual investments we really have to think about where you are on the track in advance this is a photo from Bathurst showing from this area you can actually see 20 miles out into the distance so we have to know you're gonna see 20 miles out how are we going to show this we'll do this with matte paintings we'll do this with LEDs but we have to plan this out in advance this is another shot of Prague so we know we've designed this racetrack so we know this is an area where we want to show off the whole city we know that you're going to be looking far out into the distance but we want to make sure that we're not that we have the room for this performance so we designed this so that there's almost no vegetation there's very little grass

there's not much blocking your view we're not spending a lot of attention there because we want your focus to be in the distance we thought a lot about how we were going to LOD the city we drew a lot of matte paintings for this so this type of thing takes a lot of pre-planning and design in advance if you don't go in with this sort of pre-planning you can wind up and put yourself in really bad spots and have to spend a lot of time redesigning and optimizing this just shows that we look at these every day and we're trying to turn them green that's basically office traffic graph shows we spend about two years doing this though so it's part of the challenges of being an artist in the games industry again charts of GPU probably less interesting for the group debug menus less interesting and picks so we'll just skip through all that these are all tools that we use to evaluate a scene so some of the pitfalls that we had for for all of this is we do a lot of this tuning by hand and so what we're doing moving forward for us you know in case you guys

are interested is we're looking to automate a lot of this so instead of the artists hand tuning we create levers that allow us to do things like turn up scene density turn down grass to give us more levers to play with and less hand tuning of assets to optimize for performance this is just showing you some visual examples of how a game moves through its development this is showing Rio early on doesn't look that great also has a lot of problems in performance that number in the top right is 8.8 milliseconds this needs to be down closer to round five this is after we spent a great deal amount of time optimizing you can see we've done a lot of things that we talked about composition palette value and we got the scene to about 5.8 milliseconds which was able to run just fine and these are all the different types of things that we did to get there this is another scene this is a scene that we got back right from our overseas vendors it had a lot of issues with everything we talked about again value scene composition and

artist come in and they balance it and so you can see the big the big difference that we have there and also optimize it we went from six point seven milliseconds to five point four so this is this is what we get back from our artists and then this is what we have to do to polish internally to to get it to a place where we can ship it so a big difference there and again this is calling out some of those things in specific detail showing air let me pause it real quick this is actually showing a really interesting things when we went from five to six we actually added a number of features like night and wet and a bunch of things that actually ate performance so what we had to do is we had to draw the same level prog with actually less performance than we had so we had to make it look just as pretty but we had to find clever ways to rebalance our perform our scene to deal with not having as much performance [Music]

[Music] sorry let me go back so that might be a case where you say butBut Matt five looks better and honestly for this particular scene I'm just glad doesn't look worse it was really difficult to get down but we had to do a lot of clever things with composition removing trees doing all that and this is the type of thing that artists do often to to get where we need to go this next thing and we're almost at the end here is to show you a time-lapse of how we build up a scene over the course of development but I thought you guys would find interesting so this is going to show you Dubai and Rio [Music]

[Music] yeah so just to summarize this is really again this is doubling down on core our principles and those are composition adding fidelity and detail where the players and at natural I flow is those core principles of value pallet shape and composition don't lose them we do often in the industry they're they're really sort of fundamental but it's amazing how often we forget build your scene from big to small don't focus on the fine details to quickly design the cadence of your experience design a rhythm and flow don't do don't edit everything at eleven plan for your pinch

points in advance really do a lot of work on pre-production understand what your scene is going to be before you edit check on performance daily in our case and then give your artist the tools they need to solve those issues so that's it that's everything I hope you guys enjoyed the talk and I'd like to open it up at this point to QA thank you thank you so much now that's this amazing even though I've seen been plated just seeing the breakdown is honestly it's pretty ridiculous [Laughter] there have been some questions during the presentation I kind of collected them up and I'm not in any particular order but I'll try and throw some at you now yeah we'll go from there so I thought it's a pretty interesting question do you hand draw all the concepts or they 100% computer-generated so we have we actually have a group of very talented CGR painters that they do

a lot of hand painting however they also do photo bashing it really depends on the the goal of the concept is to get the point across in as little time as possible so little use whatever tools are at their disposal to get there but there are actually very talented painters and a number of what they do is also hand-drawn okay another question how long does it take to do one track maybe like let's let's talk about like an original track for you guys how long will it take yeah that's it that's a really interesting question because it really depends on where we are in the project usually the first track that we create the you know will start will start iterating a typical development cycle for us as two years and we'll start a track relatively early on maybe six months after pre-production and it takes us about a year and a half to make that first track after that our crown jewels what we call our fictional

environments they'll tend to take maybe less time a year maybe even nine months but fictional tracks just need more time iterating our real-world tracks when were first developing them take about a year and then we get those down to six maybe even four months at times and really a lot of that is our artists are getting more familiar with the system our tools are getting more developed so it goes from a longer time to a shorter time over our development cycle so I would say on average it takes about a year to develop a track just curious - how many like how do you guys handle the art team like how many artists I know you guys also do outsourcing what what what does that kind of look like yeah so we have sort of three categories we'll have our internal artists and that's a smaller team we've actually grown over the last few years but that's typically about eight to ten on say the environment team I'll use them because it's a typical ratio for us

internal artists maybe less maybe closer to six and then we'll have maybe another 12 to 18 contractors that we also have internally at the studio at our peak and then we'll outsource to you know three or four different outsource companies and in that case we could have hundreds of artists working at it at a given time so that's that's kind of the breakdown for our environment team and that kind of scales across all of it that's our largest team but it scales across to the other teams okay there was a question about how much overall this 3d scanning have in the pipeline so it's we leverage it heavily so we do two things we laser scan which is where you put a device down and you scan in the three six two degrees and they use that laser scan we actually use it as a base we'll break it into multiple pieces and we'll send that laser scan to our vendors and they'll use it as reference when they're building out the actual asset we also

heavily leveraged photogrammetry the more we've learned about photogrammetry the more we use it and so now a great deal of our environments are our prop assets are actually built with photogrammetry so that's not laser scan but it's it's photo scanned and so a good deal of our assets are built out that way I'm a technical question how do you handle culling how difficult is it to make the collision yeah both those are good questions so we use a visibility system for calling that says because we know you were on a racetrack we follow a R we have a camera follow a spline through our racetrack and we basically set a pixel limit so we say anything that's lower than this threshold say six pixels or three pixels doesn't get drawn and so we've what we found is that we balanced the amount that we feel like we can we can set that number two without objects popping in and so we drive

around the track the camera sees what it sees and then it automatically calls out any of the objects that it didn't see from that particular area and sometimes we'll have to go in and say you know this object popped in too late or it popped into early and we'll have two handed just that but that's how we call out objects is with this camera based system and sorry what was the other question well this had to do with clicking and calling oh so for collision so we built we just build our collision from the visual mesh so often we'll loft out something like a our road and our shoulders and our barriers and then we'll duplicate that mesh and use that meshes as our collision point have you ever made an underground trap the closest we've made is tunnels small tunnels but we've never made a completely underground track but we always talk about that I think underground on the moon and under the

ocean come up pretty often but I feel like that's not likely not in our future all right this was interesting have you ever thought of using design to make it more challenging like a giant billboard to distract hmm visual design we don't usually we spend a lot of time trying to make it easier for the player so we try to we focus most of our efforts I would say in the opposite to say how can we minimize distraction to make it so that your gameplay experience is as focused as possible so no I would say that we try to avoid that great I just saw an interesting question pop up from open-world design versus linear level design any thoughts yeah so I mean it comes with its own unique challenges and we don't have to you know think about the open world too often but we are our

sister studio playground has to think about it all the time so it definitely comes with its own challenges it has its own visibility system because you can drive anywhere you want in the world you have to deal with your own collision issues you're not as focused in terms of not knowing where the player is going to be so it's harder to design moments for so what you wind up doing is you end up drawing these iconic or creating these iconic set pieces so the moments kind of create themselves no matter where you are in the world so no matter which way you're looking you're sort of seeing an iconic area so it's just a very different design challenge those guys are really good at it they do a great job but yeah I mean it's a whole subject in itself really we've talked about it quite a bit and they both create their own pros and cons um Fogg detailed even humans how many unique models of humans do you make in-game so that's that is an interesting

question early on the answer was not very you saw our crowds they're pretty terrible but the more we've developed the more we want to add you know storytelling elements the more we want to add that context for the player so recently we you know in seven we started to introduce characters we were sort of dipping our toe into it I think more and more were you know we hired an all-new to get character team we're starting to focus more on characters animation rigging creating photorealistic assets using photogrammetry so I would say that we're making a much greater investment in character than we ever have but you haven't seen a lot of that yet in the games that we've created we've just started tipping our toe into it okay maybe a couple more here so uh how do you do the clouds in this time so that's interesting we and we're actually thinking about changing that system but what we do is we capture a 360 HDR and we capture frames throughout the course of a day so we'll set up a camera rig

that captures in all directions and we'll just take take HDR exposures every say five to ten minutes over the course of an entire day and then we stitch those together using a sort of a vector-based system because we don't want to draw frames over the course of a whole day it'd be too much memory so we use a system to alert between frames kind of an a vector driven system and that produces a 24 hour day night cycle and then we use that as this sort of almost animated video as a backdrop for our environments we're actually moving to a where we're trying to move towards we're still in development a more dynamic system that at the last three clouds maybe something more akin to like Red Dead Redemption something like that sort of early days so we're not sure how we're gonna be doing that yet you know and a follow-up question to that as you're explaining is how does that impact lighting yeah and that's the

trick so what we do is we're because we have to because we have these skies we're limited in what we can do for lighting we are sort of married to whatever we captured and that's why we're looking to break away from even though visually the impact is high it's limiting and so what we're looking to do is break away and explore more dynamic ICG centered system that is allows us to light it and the scene at the same time but the question is can we get the same level of visual fidelity and impact with that system is something we're still iterating on I think that's um it's time for us to wrap up I do want to thank Matt again so much for giving such a great presentation I mean again the visuals are just stunning and hopefully everyone you you have a much higher appreciation for what into creating the visuals for this game so thank you again Matt great job and go out of thank you so I just

wanted to say you're welcome and it's been my pleasure I really like talking to you guys or showing this to you guys and ya know I really enjoy doing it