Under the Bonnet: SHIFT 2 Unleashed - 1
Slightly Mad Studios on the creation of its latest racing epic
In our biggest and most expansive developer interview to date, Digital Foundry talks extensively with Need for Speed SHIFT developers Slightly Mad Games on the genesis of their most ambitious racing game yet, discussing topics are diverse as the original title's post mortem, hopes and ambitions for the sequel, the evolution of the NFS brand and of course the technical underpinnings of the game itself.
Previously we've concentrated mostly on the technological aspects of the games we've covered - but with the assistance of Slightly Mad, in this piece we're able to expand the scope significantly to answer all of the questions we have ever had on how a top-tier racing game is actually made, the core assets the teams have access to in terms of cars and tracks, plus we've gone in-depth on the whole process of the racing simulation.
A great many thanks go to the following Slightly Mad staff for their extensive help in putting this feature together: Andy Tudor (Design Lead), Ged Keaveney (Lead Programmer), Tim Mann (PS3 Lead), Jan Frishkorn (Vehicle Lead), Andreas Moll (Art Director), Doug Arnao (Physics Lead), Dr Stephen Baysted (Audio Lead), Rob Dibley (Render Lead), Peter Nicholls (Render), Stephen Viljoen (COO), Suzy Wallace (Producer).
The scale and scope of this piece is such that we've decided to split this piece into two parts, with the concluding segment of the interview being published tomorrow.
Q: Let's start off by talking about the SHIFT post-mortem. With your first game complete, what was your overall assessment of the final product? What were you pleased with and what areas did you want to redress?
Andy Tudor: When I talk about SHIFT I always talk about 'the journey'. It's a journey of player's growing expectations and changing play styles, and a journey chronicling car culture over the years. SHIFT was always meant to be the next step in that journey transitioning players from the illegal street racing of Underground and Carbon and that 'Fast and The Furious' mentality, through 'legitimate' racing in ProStreet (that brought 'organisation' to the racing), and finally onto bona fide circuits for the first time.
We never saw Need for Speed SHIFT therefore as a single-release product even from the initial concept stage; each iteration is specifically designed to be a further step on that journey that takes us to a place where we're the most authentic, realistic, visceral racing experience out there.
So looking back on SHIFT we were happy that we'd achieved the specific goals we'd set for that particular product - we had a fantastic cockpit experience that really rebooted the pure raw fun of racing a car at 300km/h and we shook up the traditionally sterile sim genre with a number of second-to-second, minute-to-minute, and meta game objectives and rewards. The quality level was high and that was reflected in both sales and scores.
When I talk about SHIFT I always talk about 'the journey'. It's a journey of player's growing expectations and changing play styles, and a journey chronicling car culture over the years.
Therefore we knew we had a great base to work on for the sequel and the telemetry and community feedback really reinforced the roadmap we already had in terms of the features we knew we wanted to add and improve, and the areas we knew we were going to streamline or remove.
The two main areas were authenticity and variety; we knew we wanted night racing - it provides variety and we wanted to add gameplay challenge there based off the personal feedback from real-world racing drivers. We knew we wanted to keep on top of the cockpit innovation since competitors would no doubt start copying it. We also knew we wanted to mature the product even more with a premium presentation, a streamlining of the different currency systems (XP, Stars, Cash), and the inclusion of authentic real-world boss drivers and licenses (FIA GT3/GT1).
We then set out to double the number of cars and tracks - but ensure that there was variety within them (so no 'five variations of the Toyota Corolla' but categorising them into Retro, Modern, Muscle, Drift etc). Finally we then wanted players to share this experience with others - hence Autolog (which we called "The Driver's Network" at the time before discovering Criterion were coincidentally developing a similar system) which is a game-changer for the sim genre and other game genres can benefit from.
Q: Probably one of the most impressive improvements is in terms of the sheer scope of content you've added. What were the challenges here, and as a studio, how did you gear up to handle this?
Andy Tudor: At the start of any game (as most teams do) we create a wish list of what we want to achieve and see in it. This can be features, cars, tracks, anything. We already had some core features down from the over-arching 'roadmap' list and we combined this with our own wish list, the community feedback, and telemetry from the first title to get an uber-list.
Now, of course, that list is very long and not focused yet but regardless we then plug that in and see what kind of schedule we'd need to complete it. Does it fit into the time we have available? Do portions of it complement each other and would therefore work as a nice downloadable content pack? What features do we know we can do 10/10 and which would need more time to fully realise to our high standards?
After this 'bottom up' approach we then take a step back and look at the list of features from a 'top down' perspective and start thinking about the 'message' of the game or the 'X'. In the original SHIFT we set out to create a game that really made you feel the raw experience of driving a race-spec car in the cockpit. The X there was 'The True Driver's Experience' and spoke to specific features like the g-forces and sensation of speed in the cockpit, the disorientation effects when crashing, the XP system, the career progression from 'zero to hero'; it was all about you, the driver. Having an X really ties together your feature set and provide a message you're sending to your consumers: "this game is all about X".
Going into SHIFT 2 Unleashed therefore the X of "The Driver's Battle" really helped us focus that huge wishlist into something achievable from a scheduling perspective and as time went on during development we grew and expanded accordingly as each feature or asset approached in the timeline. It went very smoothly therefore.
Q: Is it true that EA gives you a "ship when it's done" level of freedom in scheduling SHIFT titles?
Andy Tudor: We have more freedom in a certain sense; we're not tied down to the usual Christmas release of previous Need for Speed titles and nowadays there are more viable release windows - January when you can spend a bit of your Christmas cash on a present for yourself or Easter when your nephew is looking for a game to play whilst they're off school for example. These other channels give us a bit of extra manoeuvring room when negotiating how long it's going to take versus product quality definitely, but we don't have an 'open calendar' like some other companies like Valve do.
Q: To what extent does competitive analysis factor into your game design? Car manufacturers buy their rivals' new models and disassemble them to see how they work. Is there an equivalent process in creating a new driving game?
Andy Tudor: It's absolutely critical and it shouldn't even be a conscious decision as a games designer; you should naturally have a passion and love for both playing games (I wouldn't hire anyone with less than 20,000 Xbox gamerscore), dissecting and analysing them, and being pro-active and investigative about the competition.
I've played games all my life from the classic Pong all the way through ZX81, Spectrum, Amiga, SNES, PSone, Xbox and 3DS. Whilst watching movies doesn't necessarily make you a good director (Tarantino accepted), it gives you an encyclopaedia of knowledge of past experience, trends, expectations, and predictions.
I don't want to play a game that doesn't push the boundaries in some way. When I was back at Sony, Phil Harrison used to have the phrase 'the first penguin' and that meant you were either the first or the best at something and it's something that appears within the first few pages of any design doc I write... What are the features in our game and why are they either the best or the innovations? And how do you categorise them within the different pillars you've identified?
For SHIFT the analysis is very clear - other racing titles are 'car owning' games; they're about the grind for cash to then collect the car catalogue available. When it comes to the actual racing we feel they're lacking (cars never deviating from the racing line, unrealistic damage, lacklustre sensation of speed, feeling of 'loneliness' when driving due to a lack of atmosphere etc) so those are the areas we continue to pioneer in; the second-to-second core gameplay rather than the menus around it. The Helmet Cam, Night Racing, and Autolog speak specifically to that and are all either the best or first in their category.
Q: SHIFT was marketed heavily as a Need for Speed title, yet the impression now is that you're spinning out and developing your own franchise. Was this a new direction spearheaded from you as developers, or was it a corporate decision from EA brass?
Andy Tudor: Again, it's part of the journey... When SHIFT was released, the Need for Speed franchise was splitting into different products dedicated to a particular audience so there was us, and Nitro, and World. The Need for Speed name was the glue associating them all.
Starting with Hot Pursuit, the Need for Speed name has now been abbreviated to the 'N' ident you see on their box art and then for SHIFT 2 Unleashed we continue on by simply using that symbol both for brevity (since "Need for Speed: Shift 2 Unleashed" is a bit of a mouthful) and also to move towards defining the SHIFT experience as a franchise unto itself. The parallel here would be EA Sports = Need for Speed and Madden = SHIFT. It was a mutual and strategic decision.
At the core SHIFT 2 has a detailed mathematical simulation of the physics forces involved, and layered on top it has the first-hand expertise and knowledge we get from real-world drivers and our own track day experience.
Q: In the console racing sim market, Forza and GT are dominant. To what extent have the SHIFT games been designed to directly challenge these behemoths?
Andy Tudor: Forza and GT are on their third and fifth iterations. So there's clear brand loyalty to those titles, a considerable consumer base, and a high benchmark in terms of quality and expectation. So we know the audience we want to reach and we know the standard required by those guys right off the bat.
Our attitude isn't one of copycatting though - you'll always be playing catch-up if that's the case. Instead we focus on questioning every aspect of the racing genre as a whole and ask why they're like that? Are they still relevant, and can they be improved or rebooted?
Elements like the cockpit view in the first SHIFT or night racing in SHIFT 2 Unleashed are prime examples there. Both have been in games for years but one always felt really restrictive and the other never gave any additional gameplay challenge. So from a design perspective those were things to 'fix' and keep the competition on their toes.
Otherwise though we don't see them as a threat; they have very clear visions for their products as do we. We're not here to play a numbers game on the amount of cars or tracks we each have as the indication from players is that that's not a high priority anyways. I've said before it would be like Call of Duty and Medal of Honor having an argument about the number of guns in their games versus the actual experience of firing a single one.
But we will continue to add the features we feel are compelling and pioneer features that the community requests (and we have a strong list on our whiteboards already).
Q: Gamers have described SHIFT 2 as being perhaps more to the arcade end of the sim spectrum up against Forza/GT, and talk warmly of a greater sense of speed. Would you concur with this?
Andy Tudor: We approach the SHIFT franchise from the perspective of the gamer - the one who grew up with Need for Speed and is now a little bit older looking for something more mature. They may watch NASCAR or Formula 1 but the sterile elitist nature of Forza/GT isn't appealing to them; they're the guys that might have moved to Project Gotham when it was around. So they're after a kickass racing experience they can play with their friends that is fresh and contemporary but moves into a more authentic area; that's SHIFT.
The design of the first SHIFT therefore brought a brand new attitude and direction but the heritage of Need for Speed - the traditions that those players have become familiar with over the years - still remained like Nitrous and 'In The Zone'. We then built upon that with the Star system which rewarded you for completing objectives during the race (so even if you came last you still accomplished something and could progress) and the XP system which rewarded you for your natural driving style and provided a system to deliver new cars, cash, paints, vinyls etc.
In direct comparison to Forza/GT, yes of course these will look like 'arcade gimmicks' but without them you would have a very dry experience - one that simply provided races to enter and money to win. And unfortunately, that's what players have been accustomed to in those titles.
When it came to the on-track experience we very consciously provided all the hooks, dials, sliders and modes to allow you to customise the difficulty and handling to your driving style since we knew that both the core Need for Speed community (who were used to a more 'arcade action' experience), and the 'sim' players needed to have that accessibility. Out of the box we evaluated you with a test lap to aid you in the decision-making process but throughout you could tweak and tailor it how you wanted. So if you wanted a more 'sim-like' experience the tools were all provided but equally your 12-year-old nephew could get enjoyment out of it opening it on Christmas Day.
The actual raw feeling of driving however wasn't weighted one way or another though; it's an emotional simulation of the feeling of racing a car at high speed.
At the core it has a detailed mathematical simulation of the physics forces involved, and layered on top it has the first-hand expertise and knowledge we get from real-world drivers and our own track day experience. There's no deliberate decision to 'dumb it down' here at all - the maths are accurate and the emotion is authentic. If it feels raw and fast and fun when racing then you have to question why you feel other titles are 'more realistic' when every tool we have at our disposal is saying otherwise.
Q: Forza and GT are 360 and PS3-specific of course, while SHIFT targets all the HD platforms. How does this add to the technological challenge?
Tom Nettleship: It was important to us that we achieve an equivalent game experience across all three platforms. While we do use a lot of platform-specific code to maintain a high frame-rate on the consoles, our dedication to cross-platform equivalence means that a feature we'd only be able to implement on one of the platforms couldn't be included. The only major exception to this was anti-aliasing, where we used the SPU-based MLAA approach on PS3, and more traditional MSAA on 360 and PC.
Q: Handing in a technologically advanced game these days requires an ability to get the most out of the unique architectures of the Xbox 360 and PS3. Both SKUs are very close to one another - what's your approach in dealing with the two consoles?
Tim Mann: We basically looked at what the required output was - be it damage, motion blur, HDR etc, and then at the initial source data for that stage. How we then got from one to the other was left to the platform-specific teams since (due to using completely different techniques on each platform) certain things took longer to process on one than the other; damage for example is a breeze on an SPU but takes longer on a 360 core. On the other hand, the 360 has a much quicker GPU in general so some graphical items had to be processed on SPU on PS3.
We didn't really impose any restrictions on the techniques that could be used, be it SPUs on PS3 or threads on 360 - just use what you have available.
For modern cars the manufacturers provide us with CAD data, images from official photoshoots, car specs, technical details, and a whole other bunch of information; anything we can get our hands on, basically.
Q: With 140 cars in the game, that's a colossal amount of work even without factoring in any of the other content. What is the process of recreating a car within your game? What tools are used and how long does it take?
Jan Frishkorn: For modern cars the manufacturers provide us with CAD data, images from official photoshoots, car specs, technical details, and a whole other bunch of information; anything we can get our hands on, basically. The car is then completely remodelled to reasonable game specs using the CAD data as reference for accuracy.
For older cars however there is far less information available and usually no CAD data at all. These cars are created purely from blueprints and photos, and therefore need highly talented modellers and deep research in order to recreate a complete car with the same attention to detail as a modern one.
Car production starts with modelling the high detail LOD (level of detail) exterior of the car, followed by the first-person cockpit. After this, the textures for both are then created and the fine details are added: realistic windscreen reflections in the cockpit, interior animations (vibrating mirrors and hoods for example), animated parts (rear wings, popup lights etc..), individual drivers, realistic multi-function dashboard displays and gauges, gearstick and shift animations, individual HUD gauges, authentic racing liveries, full crash damage setup for the car body and a pretty complicated setup for all the available upgrades and customisation options. Obviously, these features all take a lot of time to get them to perfection.
Last, but not least, the individual car physics are added...
Doug Arnao: The physics model is a full 3D scope engine capable of creating the car dynamics based on parameters taken from the specs of the real-life car - those provided by manufacturers directly and those provided from extensive research. Chassis, suspension, aerodynamics and tyres all generate their forces in 3D in real time. We run these models and recalculate the car state at 400 times per second which makes for a very convincing experience as nothing is left to guesswork.
The chassis model in itself is pretty extensive (approximately 150 parameters). Basics include weight and CG heights and positions and we model the weight and inertia of the sprung and un-sprung masses separately along with spin inertias of the un-sprung items (wheels, brake discs etc). There are 3D suspension geometry parameters for all inner/outer points of a double wishbone along with the tie-rod and damper placements.
The dampers themselves have slow and fast parameters (n/mm/sec) with adjustable switchover points and bump stops with their own stiffness settings. Differentials are the standard 'Salisbury' type with accel/deccel lock settings and visco electronic diffs are also modelled along with inputs for spring rate. Brake torque and brake heating are monitored on a per wheel basis in order to get the heating right for brake pad fade and disc glows.
The tyre model is based on the 'brush model' slip curve generation. Core parameters here include: cornering/braking/self-aligning stiffness, load and camber sensitivities, heating parameters, rolling resistance and base grip of the rubber in longitudinal/lateral directions. The tyre model therefore is class leading (if not the best) in the sim industry.
Finally, the engine model uses a standard rpm/torque curve lookup in 250rpm increments. Some parameters are: rotating inertia, accel friction drag, deccel drag, heating. Turbos are modelled as separate components with their own separate physics and can be bolted on to directly effect it as in the real world.
Once the physics are input, intensive testing takes place and the upgrades and final audio are balanced. All in all the complete production of one car takes two months (dependent on the number of visual upgrades and complexity) with usually two artists working on it from start to finish. The tools used are Autodesk 3DS Max for the 3D work, Photoshop for textures and our own proprietary toolchain for exporting, physics input, and functionality tweaking.
Q: In recreating the circuits, what kind of raw assets do you have access to? How do you go about making a track as mathematically precise as possible? Would you ever consider tweaking reality to make for a better playing game?
Andreas Moll: For existing tracks we mainly use GPS and CAD data which contain the track layout/width/elevation, armco placement and style, and gravel bed/kerb/tyrewall positions. Because many tracks are constantly changing in real life however, we then proceed to get the very latest reference data from photoshoots and research also in order to ensure the most up-to-date version of the track is recreated meticulously.
Of course we set out from the beginning to be as authentic as possible. However, what we found is that what is 'mathematically' correct in-game doesn't always necessarily 'feel' correct when playing due to the different field of view you are viewing the action from. Take Eau Rouge at Spa for instance - one of motoring's most iconic stretches of tarmac. Originally we input the CAD data and modelled the elevation change exactly as it is in real life. But when we came to play it it just didn't feel like the terrifying climb that it is in reality, mainly because you don't get the same physical feedback you have racing it in real life when lounging at home playing on the sofa.
In reality Eau Rouge is brutal - hitting the bottom of the climb makes your stomach lurch and your neck compress - all sensations that tell you of the elevation change and that can't be conveyed to the player via Dual Shock. So similar to other areas of the game where we simulate physical experiences (g-force head movement, crash dynamics) we enhance the technically correct data in order to recreate the real-world sensation. So in this case, we increased the elevation change, playing around with differing values until the game yielded the true feel - making Eau Rouge back into one of motorsport's most demanding and exciting corners. Of course, such enhancements are used sparingly only for cases where the 'emotional' experience is not being recreated by the maths - the 'Fuchsrohre' at Nordschleife and Bathurst for example.
We always start with real data that the car manufacturers give us and then plug it into our physics engine to give us a solid base. This is non-negotiable - there's nothing to interpret here or opinion to give, this is accurate to the manufacturer-spec.
Q: SHIFT 2 has received a lot of praise for the car audio in particular - what was the approach here?
Stephen Baysted: Slightly Mad Studios as a company is renowned for its fanatical attention to detail, its dedication to realism and accuracy, and the audio team is no exception.
Our objective was simply to make the player's sonic experience as immersive and realistic as possible. The complex process begins with recording actual cars - either on a circuit or using a chassis dynamometer with specialist microphones. Once these have been carefully mastered, the in-game sound modelling starts; we model every single sonic signature of each vehicle including the engine, transmission, exhaust, gearing, turbo and superchargers, surface-dependent tyre noise, tyre type, wind, chassis resonance and suspension noise. For racing cars, additional techniques are used to capture and then convey the extreme amplitude and dynamic ranges involved.
Q: On the one hand you have real-life driving, on the other you have your simulation. What is the process of ensuring that your game is as close to the real thing as possible, while retaining the essential fun factor? Did you test drive every single car you put into the game?
Andy Tudor: I think the devil's in the details - SHIFT is a 'racing' game, not a 'driving' game. Therefore the feeling you may have had screaming down the motorway or hurtling down a particularly nice country road in a stock/factory car are very different to competitively and aggressively attacking a racing circuit in an upgraded/race-spec vehicle. The sounds are all different, the acceleration is much more intense, the weight of the car is lighter but the grip is better, the consequences for pushing too hard are dire, and there's a focused mental state you need to be in whilst your body is being punished by continual movement and adrenaline. This is the experience when you actually go and do it yourself in race-spec vehicles - which the team does at multiple points during the development process.
Simply playing other games or watching it on TV is not an option here as they will both give you false readings.
We always start therefore with real data that the car manufacturers give us and then plug it into our physics engine to give us a solid base. This is non-negotiable - there's nothing to interpret here or opinion to give, this is accurate to the manufacturer-spec. So you know when you're racing that it's the real deal. Through the process of creating the car, knowing what the engine is capable of, the sounds being recorded and getting out there to race it in real life the whole picture starts to be built up. Then we put it in the hands of real experts - both the insight of real-life racing drivers and community days where players get hands-on sessions.
Testing and iteration and further hands-on sessions occur using multiple platforms and gaming setups (wheels versus pads versus the infamous D-Box chair versus multiple-monitor setups etc) to get it to a place where we're confident we're delivering an authentic yet accessible experience.
It doesn't stop at cars either. Again, that idea of an 'X' helps here when it came to the feature of 'track degradation' and how the circuit gets chewed up over the course of a race. On TV, an incidental detail like the chunks of tire rubber thrown up by cars hardly ever show up (even in HD) but the feedback from the drivers was that those things are extremely prominent - and significant - to a driver since they have the potential to lose your grip on that portion of the track. When drafting an opponent in front too, they often bounce off the hood and hit the windscreen with a particular sound causing a distraction.
This kind of feedback is invaluable therefore since it gave our track artists a deeper sense of how prominent the black tire marbles should be against the gray tarmac compared to in-car and TV footage also being used as reference. It also gave a neat gameplay bonus too - now there was a risk/reward mechanism for drafting an opponent. Sure, you'll get a minor speed advantage but there's also the potential to be caught in a wake of his tires kicking up both gravel and tire rubber meaning a distraction as they hit the windscreen and a potential momentary loss of grip.