Detroit: Become Human is a different kind of tech showcase

As big budget triple-A games fixate increasingly on delivering larger, more complex open worlds, we’re left wondering – what if all that power was concentrated instead into smaller scale environments with a focus on extreme detail? That’s exactly the approach we see with Detroit: Become Human, with developer Quantic Dream delivering its best game yet – and a polished, intricate presentation quite unlike anything else seen on the market today.

Detroit is built on the latest iteration of Quantic Dream’s in-house engine, and it’s the first original PlayStation 4 title released by the studio. However, its concept is rooted in a 2012 technical demo presented at GDC known as “Kara”, which was designed for PlayStation 3. This demo serves as a teaser for what would become Detroit, while acting as a reference point of sorts for how the technology would evolve over the coming years. It still looks good but the final game is a significant leap beyond that initial demo.

At its core, Detroit shines brightly on both the PlayStation 4 Pro and the original PlayStation 4 console. When using a Pro, Detroit makes use of checkerboard rendering to reach a 2160p pixel count, but many of its post-processing effects are rendered at lower resolutions to save on performance. On the base system, Detroit instead offers a full 1080p image. Both versions make use of an extremely high quality eight-tap temporal anti-aliasing solution, in order to eliminate edge shimmer and in-surface aliasing – Quantic Dream reckons the fidelity holds up against 8x MSAA. It requires just over 1ms of processing time from the game’s 33ms per-frame render budget, making it an effective and fast solution.

In practice, the game’s focus on post-processing actually makes the difference between the two more difficult to pull out in motion – there can be variations in aspects like volumetric light resolution, but side-by-side, both versions look extremely similar. Based on what we’ve played so far, PlayStation 4 Pro’s biggest advantage seems to be performance-based – there are occasional fluctuations under the 30fps level on both systems, but the Pro generally loses fewer frames, while the taxing more open areas see a more profound advantage for the enhanced hardware. The bottom line is that resolution isn’t a crucial component of the presentation overall, and both PlayStation consoles deliver a beautiful look.

Quantic Dreams’ objective was to create an engine that could support a variety of unique environments with lots of dynamic lights, along with variable weather conditions such as rain and snow and advanced direct and indirect lighting. According to the studio’s GDC presentations this year, the latest iteration of its technology utilises clustered forward rendering – an approach that offers many of the advantages associated with forward rendering including a single pass for geometry while also handling many dynamic lights. Standard forward rendering requires lighting calculations to be performed on every visible vertex and fragment, making it computationally expensive to render lots of dynamic lights. By breaking the scene into clusters instead, however, it becomes possible to render more lights within the strict render budget.

While exploring the game, there’s a genuine sense of realism on display but just how is this achieved? It’s a complex question, but one key element lies in its materials systems. Detroit makes use of physically-based rendering to properly simulate roughness and reflectivity of light across materials. When light reflects off surfaces such wooden floors, fabrics or drywall, it dissipates across the surface, while shiny metals or wet pavement present tighter reflections – just like real-life. And all those little imperfections and microscopic abrasions present across real-life surfaces? The utilised BRDF (bidirectional reflectance distribution) ensures that these accounted for. Light also respects the laws of physics thanks to partial energy conservation – which basically means that the reflection of light cannot exceed the brightness of the original light source. All this is to say that like other PBR models, materials are designed to react naturally and realistically to incoming light sources.

Our detailed 20 minute breakdown to Detroit Become Human’s exceptional technology, with PlayStation 4 Pro as our focus.

Every environment in the game is lit using a mix of dynamic lights, which are lights which can be moved around the scene, and baked global illumination data. The question is – what does this mean for the player? Put simply, lighting looks fantastic. The game sends you through a wide variety of areas and this approach to lighting boosts realism across the board, while allowing Quantic’s artists to present even more striking scenes. When exploring Todd’s filthy hovel early in the game, the area is predominately lit indirectly, but it appears strikingly realistic in practice. Other areas – those set outdoors, for example – appear far more vibrant with bright lights and reflections everywhere.

By limiting the scope of the game to specific key areas, however, Quantic Dream can focus more of its time and effort into creating much richer scenery. Many of these features are present in other game engines but it’s the coupling of excellent lighting with carefully modelled and textured scenery that makes all the difference. There are certainly open city areas in Detroit and they showcase the engine at a larger scale, but it’s often the more confined spaces which impress the most.

Another key element in the presentation is character rendering – after all, this is a story-driven game and you’ll be seeing a lot of close-ups of each major character in the game. Detroit makes use of extremely high polygon character models that are all superbly shaded, textured and animated. High quality models have become increasingly common but even then, Detroit reaches new heights.

When examining a digitised actor, we tend to focus first on the eyes – and Detroit offers a good solution here with a bespoke eye shader and movement mapped to the performance capture data. As a result, in most scenes, the eyes no longer have that dead look you may remember from Heavy Rain. Even more interestingly, there is a unique, perceptible difference between androids and their malfunctioning counterparts, dubbed the ‘deviants’. When the androids are acting within their programming, there is a sense that the eyes of these machines appear slightly less realistic than the humans around them. As they take on more human attributes, however, it seems to slowly change.

Then there is the surface of the skin. A screen-space sub-surface scattering technique is used to simulate light penetration across the skin and is used heavily throughout. This allows light to penetrate and dissipate throughout the skin and fat of the digital actors. When you hold your palm up to a bright light and see a pinkish glow around the edge of your hand, that’s subsurface scattering in action. Hair is another key element which is always difficult to render but the selected technique generally produces excellent results, and Quantic Dream has also made a good effort here in animating hair depending on conditions. This extends to cloth rendering on costumes, which possesses a realistic simulation model reminiscent of Assassin’s Creed Unity. It looks excellent in motion. The dedication to achieving a realistic look even extends to the mouths and teeth of the characters – again, the lighting is spot-on and remarkably authentic.

Cinematic sequences are further enhanced by a combination of close-up lighting and shadows. When engaged in conversation, secondary lights are used for each character, enabling them to stand out in each scene – like using a studio light while filming a subject in real-life. Objects within a 10 metre radius also receive an additional close-up shadow, which basically means that objects within this area produce cleaner, more defined shadows. This helps avoid the typical shadow artefacts you might otherwise encounter.

Character models have always looked great in Quantic’s games but it’s the animation and performance of each digital actor that really stands out. The studio uses a full performance capture rig to bring the actors and actresses into the virtual world to great effect. The team first utilised this with Beyond: Two Souls but the more realistic rendering available in Detroit increases the realism significantly. These characters are further highlighted by the excellent bokeh depth of field – basically, the circular shapes visible in out of focus regions. This is a natural occurrence that can be produced using the right lens with a camera and the shapes presented here add to the cinematic presentation.

And really, with Detroit, it genuinely feels as if the developers were focused on simulating a real camera within a virtual world. The concept of ‘interactive cinema’ has been around for many years, as has post-processing designed to simulate camera behaviour, but Detroit goes one step further. It genuinely feels as if it were shot more like a film than a game. The angles, cuts and camera tricks all manage to create this feeling and depth of field is just one element of the presentation.

fps2
By and large, both PS4 and PS4 Pro deliver a fluid 30 frames per second. However, in stress points – like large, open city scenes – the Pro’s small advantage becomes somewhat more pronounced.

Per-object and camera motion blur are other key components used to further the sense of realism. The intensity of motion blur in filmed footage varies based on the selected shutter speed of the camera and virtual shutter speed must be considered when replicating this effect in a game. Detroit uses an agreeable shutter speed that looks suitably filmic, but manages to avoid obscuring the detail present in the image. Like most post effects in the game, motion blur is rendered at half resolution to…

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *