spatialcomputing.info

In the decade that I’ve spent studying and reporting on VR I’ve had the fortune of experiencing a wide range of VR game design—a craft which is still far from ‘figured out’. And while I have no intention of becoming a game designer by profession, I thought it might be a good learning exercise to try my hand at designing something in VR to better understand the challenges and opportunities therein. And here I’d like to share a bit about what I learned through that process.

For context, this was a personal project conceived by myself and my developer friend Henry Ventura. We set out to see how effectively we could translate a weapon from a non-VR game into a working VR version that maintained the weapon’s ‘character’.

We looked to the arsenal of Respawn Entertainment’s Apex Legends (and its cousin Titanfall) for inspiration. The game’s iconic weapons are all quite beautifully modeled and animated, giving us a great blueprint of both function and ‘character’ to work from. We chose the Wingman revolver, and here’s what we came up with in less than a day of development time.

This was built with Unreal Engine 4 running on PC; unfortunately we won’t be distributing a demo build.

What I Learned

Although I helped a bit with textures and sounds, my role in this project was primarily interaction direction—that is, directing the way the weapon, and its various interactions, should feel.

Choose Your Weapon

The first part of the process was, of course, choosing which weapon we were going to build. If you were building an actual VR game with guns, the type of guns you choose to use can massively impact the rest of the game because of how their interactions will dictate both the player’s attention and the game’s overall pacing.

The choice of the Wingman revolver was highly influenced by its one-handed operation. While two-handed weapons can definitely be fun in VR games, they also demand more of the player’s attention and result in the player having one less hand to interact with the environment around them.

With the Wingman in particular, the reload gave us a great bang for our buck because the player gets to use both hands for the reload without needing to dedicate the off-hand entirely to the weapon. It’s clear that Valve’s Half-Life: Alyx was also built around this idea, as the game features only single-handed weapons—even going so far as turning a shotgun (pretty much always a two handed weapon) into a shotgun-pistol.

Character Acting

A major goal of this project was to bring the Wingman’s essential ‘character’—the way you imagine that it would feel—into VR. If we were building from scratch, we’d need to spend time figuring out what that character should be. In this case, Respawn already did much of that heavy lifting by giving us a very clear vision of the gun’s look and feel through animations and sounds that we drew upon.

I was actually quite surprised how effectively we were able to bring the character of the gun into VR. The Wingman is a hefty gun that feels like something an expert gunslinger would confidently use. Our finished VR version, especially with physics-driven gun-spinning as an optional flourish, really does make one feel like a badass gunslinger.

Our success in effectively translating the gun’s character into VR is interesting in itself because it suggests that designing VR weapons in the ‘traditional way’ (ie: by animating how you’d like them to look and feel) could be an equally valid starting point for figuring how the weapon should behave in VR. In the traditional method, it’s an animator’s vision which defines the feel; that artistry, it turns out, can act as excellent ‘concept art’ for how a gun should be designed to feel in its VR incarnation.

A carefully designed weapon in VR can make the player themselves feel a certain way. It isn’t just about pointing the gun at enemies to make them disappear. A VR gun can be almost like a ‘costume’ that gives players license to ‘play act’ in a way encouraged by the game designer.

A little secret… the fan-firing seen in the video (when the off-hand fires the gun by pulling the hammer back) isn’t a real mechanic, it was the result of ‘acting’. We came across the idea while literally playing around with the gun and acting out a fantasy.

When we realized how fun it ‘felt’ (even though we were pretending the mechanic existed), we knew we had to include it. Since this was prototype, there was no need to build out the actual mechanic when it could simply be ‘performed’ to get the point across. Fake it ’til you make it.

Communication During Development

While we were developing and testing the Wingman interactions, I found it was often difficult to articulate exactly what kind of changes I’d like to see. I take it this is normal for all game design, but it seems especially true for VR given the immense freedom that the player has to interact with objects.

In many cases one of us would be using the Wingman in VR while the other watched via screenshare. This work reasonably well to try things out and talk about them on the fly, but at times I was wishing that we could simply both be in VR and looking at the same thing together. Simply being able to point with my finger in 3D space and have my own version of the gun to (again) ‘act out’ how certain things should work, would have sped up development and made for clearer communication.

Capturing the Video was a Surprisingly Large Task That Could Have Been Optimized

I have a newfound appreciation for the Valve developer who live-captured all of the shots for the Half-Life: Alyx trailer, which really does a good job of selling what the experience feels like when you’re actually inside of it.

We really wanted to have one continuous sequence that would both demonstrate the full operation of the gun and truly show what it felt like. Between choreographing the actions and practicing the gun spins and aiming, it took several hours to get a shot we were finally happy with—and this was only a 30 second video!

Simply using the gun in a way that you naturally would in VR, and capturing the output screen, looked pretty bad because there was lots of head movement and the cropped field of view meant it was hard to tell what the framing would look like to the spectator.

Our process was to have one of us using the gun in VR while screensharing to the other. The person outside of VR would help direct each take to ensure that things were framed well and that there wasn’t too much camera movement from the spectator’s perspective. In hindsight, there’s a few things we could have done differently to speed up the process and end up with a better result.

While the live direction was certainly helpful, our biggest stumbling block was the skill needed for execution! Aiming had no assists and gun-spinning was fully-physics based, so a major portion of the time it took was simply the practice necessary to do it all with no mistakes (all while ensuring the head movements didn’t impact framing). We had tons of takes that were almost right, but we often missed a barrel or had some other minor stumble.

It certainly would have saved us time if we had scripted the barrels to register a hit with each trigger pull (no aiming required). While this risks looking bad if the aim is too far off, it would have given us a much larger margin of error to work with, and allowed us to spend more time perfecting other aspects of the shot.

Another useful change would have been to project a frustum onto the environment that matched the field of view of the spectator camera. This would have allowed the person in VR to see exactly what portion of the action would be captured in the final output, and also provided clear feedback to show how much the spectator camera was being moved when the head was moved. A smoothed spectator camera (as seen in some VR games) would have helped too.

The end-all-be-all solution would be to record the actions through the game engine so that we could make minor adjustments and then re-render the scene after the fact. This would also have given the option to render the video in higher fidelity (which would be especially useful for anyone making a production VR trailer).

– – — – –

I’m really happy with how the project turned out. Not only was the gun fun to use, but I gained a new appreciation for the challenges and opportunities of a small slice of VR game design.

While VR is capable of much more than just shooters, there continues to be a huge demand for VR games based on gun combat. If the industry is going to be beholden to that demand, I would personally love to see more attention paid to the ‘character’ and interactions of VR guns.

To me, the interactions that happen within arms reach are among the most compelling parts of a VR game. VR guns should be thought of less as laser pointers that cause damage to distant enemies and more as interactive tools.

Thinking first and foremost about how the player will interact with the gun (rather than how the gun will interact with enemies)seems like an essential part of designing a VR gun with character.

While we stuck with a design rooted in reality, this project has further convinced me that there’s a huge untapped opportunity for creative ‘sci-fi’ guns that work in totally novel ways that are uniquely fun in VR. If guns are going to be the essential ‘tool’ the player is given in many VR games, I’d love to more thinking outside of the realistic assault rifle, SMG, sniper, shotgun paradigm.