Develop asks what Vicon's new head-mounted facial capture rig means for games
Vicon has unveiled a new piece of hardware it hopes will revolutionise motion capture in games and beyond.
Keen to learn more, Develop caught up with Vicon’s hardware development manager David Reynolds, and Stuart Adcock, technical art director at Ninja Theory, whose expertise in full performance capture makes him well qualified to offer a developer’s eye view of the hardware’s debut.
What makes Cara so distinct to a peer of head-mounted motion capture hardware?
David Reynolds: The real point with Cara is that it is designed for high-performance full motion capture in a way that is slightly different from the traditional approach of using larger, more expensive multi-camera systems with optically-based reflective markers.
Typically, in the past you’d have these very expensive systems to capture body markers and face markers. And those approaches have limitations beyond just the cost and complexity; things like line of sight. So people have been building their own prototype systems for facial capture using head mounted systems for some time now, often with a single camera giving 2D output.
Cara is about addressing the market-wide need for a solution of high quality head-mounted hardware that controls software and gives users 3D. It’s important that it’s a full solution – not just for those working at the biggest studio – but for all sizes really.
So this high-end hardware really is in reach of a range of studios?
Reynolds: Yes. I really think that’s the case because it offers such a complete solution, and because there’s a relatively small investment in terms of cost and in terms of infrastructure; you get everything in one box, can lift off the lid, plug it in and be up and running in an hour.
How long has Cara been in development at Vicon? How did the project come to be?
Reynolds: It’s actually been a fair while. We started with a prototype system called the HMC in about 2008 for one customer, ImageMovers Digital, and it was used in their A Christmas Carol movie, and then the equipment went to Digital Domain for Tron Legacy, and it saw use in a few computer games and so on.
After a while we could see it was a successful product, but it wasn’t a full release product, and there were still limitations with the data then, and a lot of post-processing was needed. At that point we decided to take our time and develop it into what it is today.
And what is Ninja Theory’s interest here?
Stuart Adcock: In terms of Cara, while Ninja Theory has not used it as of yet, at the same time we have taken the journey of going down the route of developing a lot of our own tech of this kind.
I think it’s fair to say we’re one of the main game companies pioneering this kind of approach, starting with Heavenly Sword and taking it right up to today. Motion capture and full performance capture is one of the main pillars of our studio, and we’ve invested a lot of time in that space.
We’ve done a lot of R&D in terms of facial solvers that interpret the data captured to allow us to give our characters expression, and that’s been needed as until now the only option really has been using single – and therefore 2D – cameras.
Two-dimensional data in that context isn’t that good in terms of giving expression and character, meaning you have to do a lot of analysis of data and mathematics to best fit 2D data and predict the right Z-depths.
We’ve always managed to get really good results that way, but it takes a huge investment of effort. It’s worth it though, as with head-mounted cameras you don’t lose the face when actor’s look down or move close together, and you don’t have to have the density of external cameras.
Head-mounted systems allow you to do much more in the volume with more people, and we embraced that in DmC. But the chance to work with 3D data in that way is a huge opportunity for people making games.
And why in particular is 3D so important in terms of facial capture data?
Adcock: The main limitation of most head-mounted cameras to date is that they’ve been single-camera, two-dimensional images, and so you’ve needed more bespoke software. And it seems Cara’s targeting a broader target market.
What it provides is easier to understand, in that the captured 3D data better represents the 3D form of the face. Mathematically solving based on 2D data is more limited.
And 3D opens up more opportunities of what you can do with the captured data. Certainly, one goal for us is that if we can start using genuine 3D data, it can be like the old days of multiple cameras, but without the restrictions of coverage and application problems and some of the other restrictions of reflective markers.
How do you see that impacting the games people play?
Adcock: I can only really speak in terms of Ninja Theory, but our goals have always been to try not to draw a player out of a game’s world too much.
We often tell stories with characters, and we feel the best way to maintain character and give good character progression and arc is to really embrace working with actors. The best way to do that is to try and capture every nuance you can of an actor’s performance. Going forward with games, that’s going to become more normal, and things like Cara could be part of that.
You mention actors. How have you tried to make sure that Cara’s form won’t intrude too much on an actor’s freedom to perform?
Reynolds: Certainly, Vicon took on the challenge of not just providing the electronics and the technology, but of designing the head-rig in a way that balances comfort with stability. It was a significant challenge, and that’s part of why we’ve been developing this for five years.
You want to hold the frame as still as possible, so the movement of the camera in relation to the face is kept to a minimum. But the best way of doing that is probably to clamp it on to the point that is uncomfortable.
Then there’s robustness to think about, and the designing of a frame that will fit a range of head sizes and shapes.
When we’ve seen studios create their own head-rig, we’ve often seen them build a bespoke head unit, almost shrink wrapped to the head, but that can be time-consuming to put on and hard to use with more than one performer. So there were a lot of challenges in that regard, but we’ve worked hard to overcome them.
And Cara will integrate with almost any existing mo-cap environment, regardless of hardware?
Reynolds: We decided that we wanted to make a system users could dip in and out of wherever they want.
So, if you have a full Vicon system, it will work very nicely, but if you don’t have a Vicon set-up it will still work very well, and you could even use it without the rest of the motion capture system, and it works indoors or outdoors or anywhere you want.
Ultimately, the data you get out of it is something you can feed into the world of gaming relatively easily.