28 de agosto de 2018
Aspyr uses Unreal Engine 4 to bring the dark-science fiction mystery ‘Torn’ to life in VR
A dark science fiction mystery inspired by television shows like The Twilight Zone and Black Mirror, Torn offers a journey into the world of an eccentric inventor/scientist and the spooky mansion he left behind after going missing 64 years ago. A narrative puzzle game by design, the player will bring the mansion to life while reactivating the scientist’s many mysterious machines. All the while, the player is pulled deeper into the rich narrative and story building that Aspyr has worked so diligently to create.
Encountering problems that traditional development may never have to think of, Aspyr leaned on Unreal Engine 4 as it endeavored to truly make a mark in the VR landscape. We chat with Neill Glancy, Creative Director for Torn, to discover what it took to bring a VR title from its early brainstorming sessions to shipping its final code. Can you give us some background on Torn and how the project came about?
When we first started the IP discovery process here at Aspyr we didn’t have a specific title in mind to build. VR was a strong talking point across the industry as so little was known about it, so it seemed like a good direction as there was no doubt the team could learn a lot by working in such a new medium. As the Creative Director of the team, I've always felt it’s important to fuse learning new things (keeping your knife sharp) with a story or creative context the team is excited about (the wrapper). I had a Vive rig at home so the team got together over beers and we looked at some of the initial releases titles like The Blu, Job Simulator and Valve’s The Lab. This was our first experience in room scale VR and everyone was very excited about the creative space and potential.
We spent time asking the question: “How is first generation VR succeeding, and where is it falling short?” As a small team with a small budget, it was important for us to determine early where we would place our creative bets. We couldn’t support tons of features or vast scope so we had to choose carefully which elements we would define as the game’s “tent poles.” From our look at the VR market, we determined the following statements were generally true of initial offerings.
- The sense of being teleported to another reality is amazing and unparalleled.
- Games we played had a fun VR centric mechanic vs. standard console mechanics.
- Most of the early games or experiences had little story or context.
- Many games featured static worlds that were beautiful to look at, but often couldn’t be interacted with in compelling or natural ways.
From this shortlist we were able to create some desired parameters to guide the next phase of our discovery work. We thought of this as VR 2.0 and asked what would be the tent poles for such a work? The things that jumped out to us were:
That we needed to embrace the touristic VR fantasy, place the player in a world that was richly compelling, one they wanted to pick at and discover more about.
- Give the player something fun to do while giving them a story reason for doing it.
- Make the experience sticky by tying it to a “page-turning” narrative, maybe some sort of mystery or detective story?
- Make the world as simulated as possible so that it could be beautiful and convincingly alive.
- Be aware of session length in VR and try to solve for that in a way that made sense for the creative.
- Try to come up with an idea that spoke to the current Zeitgeist, a topic or spin on things that we knew people already enjoyed in other media.
Several years earlier I was the Lead Designer on John Woo Presents Stranglehold. Stranglehold used a ton of physics systems and this was an area I had a lot of experience in and wanted to leverage for the new work. In looking across the tech spectrum of physics I came across this video which I found to be very inspirational.
I was really fascinated with the water simulations and felt if we could do something with them in VR we might discover something interesting so we started brainstorming ideas where players could interact with fluids in VR. At the same time as spitballing creative ideas involving fluid simulation, we started to explore Flex in Unreal to make sure we weren’t totally insane and that we could run a fluid simulation in VR at the high frame rates VR requires.
It became immediately apparent that it was possible to render fluid in VR but only a small quantity of it. So all of our large-scale water/ocean ideas had to be abandoned. Instead, we would have to focus on giving the player a small amount of liquid and make that important somehow. We experimented with all sorts of ideas, maze games where the player controlled the orientation of the maze as the fluid ran through it, puzzle ideas where the player had to construct a solution with limited resources to get the fluid across a stage, and blob ideas where we made the fluid very gelatinous and the player controlled this roaming 1960s blob creature as it consumed small towns. At this time we made a minor breakthrough which was disabling gravity. All of our ideas had used gravity to bind the fluid to the scene but when we removed gravity the fluid was transformed into this floating tentacled “thing” that was super interesting to observe in VR and really fun to interact with and splash. Seeing the fluid splatter and twirl in low gravity was super compelling, riveting even, but the question ‘How will we use this?’ still remained.
The other challenge was that rendering the fluid simulation was extremely GPU intensive. Whatever the answer was if we wanted to let players interact with this thing it would likely have to be in some barren context as we couldn’t render the fluid and have a complex world around it at the same time.
These technical and performance constraints led to the following creative epiphany. If we could only have a small amount of fluid what was the most important and valuable thing it could possibly be? The answer we came up with was a person’s mind. The mind itself is such an abstract thing to imagine the fluid representation seemed like a perfect fit. Moreover, maybe this person's mind was trapped somewhere abstract, this would allow us to make that reality lightweight to compensate for the GPU load of the fluid.
It was becoming increasingly obvious that a story was emerging from the discovery process and technology research. Somehow a person’s mind had become trapped in some “other” place, but was that it? It didn’t feel balanced, it felt like we needed to think about the work as a swinging pendulum. One side would be this odd other realm, but what was at the other side of the pendulum’s swing? How could we make what was on that other side complement and contrast the lost mind realm? It was in this moment that the idea of the mansion came to be. What if the game’s pendulum swing was that the player spent time in the mansion and then time in the other realm? This would provide two very unique and differentiated realms for the player to experience and a natural rhythm as the games' undercurrent of progression.
We also knew we wanted to tell a story so we needed actors that could deliver a compelling performance, but we didn’t have the budget to attempt a real facial performance nor do facial mocap etc. We felt the fluid simulation could be a real win here if we could feed it the dialog and tune it in such a way that as it spoke the fluid responded. We used test audio data from classic movies to feed test lines through the fluid simulation things like lines from Dirty Harry, Taxi Driver, The Three Stooges, etc. This let us see how the fluid responded to what we knew was a good dramatic performance and dial in the parameters around the fluids response to the WAV data.
With our fluid character now looking promising, we turned our attention to the mansion itself, its architectural design and focusing more on story so we could understand the scope of work. We hired two world-class writers to work on the project to help us write the story, but more about that later.
What sort of puzzles can players expect in Torn, and what makes for a well-designed puzzle to you?
Once we knew that a portion of the game was going to be set inside a mansion we knew we wanted to give players something interesting to do there. Ideally, this thing they would do would advance the story to meet our goal of “Give the player something fun to do while giving them a story reason for doing it”. We also knew from studying puzzle best practice (thanks GDC talks and Gamasutra!) that getting puzzles right was a very iterative process. To mitigate this risk we determined that puzzles should be agnostic to the mansion’s art. So, in other words, a puzzle could be constructed, torn down and rebuilt without the artists having to do any work to support the change.
We also didn’t want to make a puzzle game that was super challenging or twitch based. It was important for us that puzzles be something that might temporarily baffle players but never block them as we wanted folks to complete the game’s unusual story. And this is where we stray away a bit from conventional game puzzle theorem. We felt that, for VR specifically, having an accessible puzzle element was more important than building an intense challenge that only a few could complete.
We also identified “hero rooms” which had the potential to have bespoke puzzles, something to mix it up a bit and keep things fresh. These bespoke rooms would leverage the technology layers we had already built for classic room puzzles but put a new spin on them. I think overall we managed to get a good balance of bespoke room puzzles and conventional ones.
Personally, I think good puzzle design teaches the player lessons as layers and then asks them to combine those layers in unexpected ways or sequences over a series of more elaborate challenges. It’s all about the “Aha” moment. Puzzles in Torn are all tied to rooms, each room has three puzzles that increase in complexity. In our backstory, the mansion is actually more machine than building so the puzzles revolve around “waking rooms up” by activating and completing circuits in that room that bring its mechanical elements to life. We also liked the correlation between completing a room and VR session length. In general, most players can complete a room in around 15-20 mins and would then travel to the other realm to converse with the fluid character. These two swings of the pendulum (mansion play and fluid realm) both fit into a 30 minute window which correlated well with what other developers were telling us about VR session length.
Are there unexpected challenges that VR presents for the genre, and how do you overcome them?
Some of our greatest challenges have been in two main areas:
Player Attention - It’s well understood that as the player can look in any direction in VR, it can be hard to tell them a story or be ensured the player is looking in the correct direction at a specific moment. To help provide player direction at specific moments we use a combination of lighting, motion, and sound to get the player's attention then query the HMD to see if it’s worked and they are looking generally in the direction we want them to. It’s a bit of a dark art and there’s still lots to learn here.
Multiplatform - The second biggest challenge was making the game perform well across the spectrum of hardware targets we wanted to hit. Different platforms such as Vive, Oculus or PSVR have greatly different levels of competency when it comes to the basics like tracking fidelity, tracking solution type, CPU/GPU and memory speed. We took time to study the emerging best practice solutions inside of other games to make sure our solutions were best in breed and something the players would feel comfortable with and be able to grasp quickly.
The narrative is obviously an essential component of Torn. What can you tell us about your approach to storytelling?
Story was one of our big tent poles and we knew audiences were eating up titles like Firewatch, What Remains Of Edith Finch, Black Mirror and Stranger Things and these experiences featured qualities we wanted to bring to our story. One of our other constraints was we couldn’t deliver animated characters, so whoever was speaking on screen couldn’t be an actual character model. We had already cast our fluid simulation as one of our central characters and we knew we wanted our player character to speak to give the experience momentum and help with player attention. Our last characters were what we call ‘Specks’ - small bright lights like Tinkerbell. These Specks were fragments of the fluid character’s mind that were trapped in the mansion and we would ask the player to collect them to repair his mind and slowly unravel the story as the fluid/blobs memory was restored. We liked this idea as it correlated with a story we had seen successfully executed in Christopher Nolan’s Memento. We felt this backbone of unearthing the story as the central character remembered it would be interesting for players and also provide fertile creative ground for player plot misdirection to keep things interesting.
This of course also made the story very much more complex than a conventional tale and it was clear we would need a professional full-time writer to help us deliver it. Initially, we worked with the extraordinarily talented Matt Soell for two months upgrading the early story draft into a true three act structure as well as answer some of the story’s bigger questions - such as “what is the twist?” It was super important to us that the story needed to be built on a very solid backstory foundation so everything in the world felt true to that world and made sense. We wanted to make a world with history that the player could poke and pick at and provide answers to any given place should they look there. As Matt’s contract came to an end we again sought a writing partner who could help us deliver the best story possible and as the fates would have it Susan O’Connor was available. Susan had worked on Bioshock, Tomb Raider, and many other celebrated titles so we were thrilled to have her join the team and help us complete our narrative vision.
The story structure in Torn follows the pendulum metaphor. When the player is exploring the mansion they have two primary activities;
- Complete room puzzles to wake up the mansion’s exotic mechanisms.
- Collect memory Specks and deliver them back to the trapped fluid mind.
This broke the narrative into two distinct cycles - one of puzzle solving in the mansion and then the reward phase of returning those memories to see what the fluid character could now remember to push the plot forward in the other realm. This pendulum-like swing between the realms felt like it had a good tempo and it’s one of the creative risks we took that I’m most proud of.
To keep the player company in the real world of the mansion, the Speck characters could talk. We thought of them as tiny mansion docents or room butlers. These Specks could tell the player the history of the mansion and its devices and rooms as the player organically explored the building but also serve as assistants and puzzle helpers should the player become stuck. They also offered a chance to add some light humor to the game world which we felt was a good match to the Mansion's somber presence.
In addition to these measures, we also created what we called “memory constellations”. During the player's visits with the fluid character, he would tell the player about things he remembered after receiving a dose of memory Specks that the player had collected mansion side. To make these story moments more compelling, we employed Tiltbrush illustrations. As the fluid character would tell the player things he remembered the player would see these recollections appear in the heavens to illuminate the story in a colorful and delightful way. This also greatly helped in keeping players engaged during the core story delivery as they never quite know what to expect during each visit that they have with the fluid character.
Why do you think mysterious old houses continue to intrigue us and how did you work to keep the setting fresh?
Old buildings hide stories and human beings are curious creatures so the two go together very well. To put a lens over this itch and make our mansion unique we also played with the idea of slapdash mechanical intrusion. In Torn’s story, the player is told that the mansion is the home of an eccentric couple who conducted weird experiments into the mind. The mansion is a container that they worked inside of if you will. Having said that, the inventor has no qualms about knocking walls down to feed oversized cables through and generally connect all of the strange apparatus together. It's a fusion of building and technology such as one sees with trees and temples in Angkor Wat in Cambodia.
We wanted these mechanical intrusions to be objects of curiosity and to look purposeful but not necessarily menacing. The term we came up with to describe this tone was “Wonk”, so we would ask the question “How would Willy Wonka build this device?” We wanted these devices to feel friendly but also mysterious. To facilitate the wide variety of Wonk installations throughout the mansion we created a Wonk part set that could be used to build all kinds of strange yet form family familiar objects.
In addition to this, our Principal Artist Chip Patterson also used to be a professional architect so his structural sensibility helps our mansion feel grounded in real-world building practicality. It’s kind of this unseen thing that you just feel in the Mansion’s DNA.
Did you have prior experience with Unreal Engine, and why did you choose to use it for Torn?
Everyone on the team had used Unreal Engine before, and I had used Unreal Engine heavily on Stranglehold. For us, Unreal was a great solution as it has powerful performance optimization tools that we knew we would need to make world-class VR visuals performant. The source control integration and UI made the tool easy to use and helped us spend more time making great content without being bogged down.
The other big win with us for Unreal was NVIDIA’s Flex and Gameworks integrations which allowed us to hit the ground running during our discovery phase.
Can you tell us about a favorite tool or feature of Unreal Engine 4 and how it has aided development?
The Blueprint system allowed non-programmers to create interactive content easily and this was important to us as Torn only had one programmer (Thanks, Alex)! By leveraging the Blueprint system we were able to audition many ideas before committing full production resources. It also had the side benefit that it allowed us to create complex instanced objects with cooked in logic that could be copied and placed around the game world quickly and then linked to the puzzles.
VR gaming is still in its infancy. What do you envision for the future?
There’s no doubt in my mind that VR is a new paradigm for man-machine interaction. The application of VR in education and industry has already proven to be strong and compelling. VR enables a new type of interaction that I call ‘spatial comprehension’. Engineers and scientists can now look at medical or scientific data in a way that’s never been possible before, is intuitive and adds great immediate value. Hopefully, the future will give us more affordable and lighter, more comfortable headsets with less setup fuss than we see in today’s rigs. The more we can decrease friction in setting up a space for VR by employing inside out vs outside in tracking the closer we’ll be to making VR truly portable and convenient for everyone.
Where can people go to keep up with the development of Torn?
To keep up with Torn news please visit https://www.tornvr.com/