Approaching the VR Inflection Point: We Need Picks and Shovels
Hold your breath: virtual reality seems to be on the cusp of a breakthrough. I’ll excuse you some skepticism, as almost every year for the past five years has supposedly been the year that VR becomes mainstream.
The sales data, however, does not lie. While Facebook has yet to release detailed sales figures for its Oculus devices, it has announced that it has sold more of its Oculus Quest 2 in five months than it has sold of any other Oculus device over the past five years. As an entry-level device priced at $299, the success of the Oculus Quest 2 shows that virtual reality is beginning to break into mainstream adoption.
My thesis is (and has been for a while) that VR is coming in a big way. Gaming has been the arrowhead for adoption, but the use of VR will expand into entertainment, education, healthcare, and many other global industries.
Growth in the numbers of headsets worldwide means a growing demand for content. And yet, as VR approaches its inflection point, we lack a content ecosystem to service the technology. We therefore need to consider the ‘picks and shovels’ of content creation — what are the platforms and tools needed by content creators to build the immersive environments and experiences demanded by the growing population of VR users?
The 3D content and graphics that underpin VR aren’t new. Computer games and animations for film and television have been the driving force behind the technology. RenderMan — the programme built by Pixar to power its CGI revolution — was described by Wired on its 30th birthday as “perhaps the most transformative piece of software Hollywood has ever seen.”
Quite a bit of the VR value chain has consolidated rapidly — headsets are dominated by Facebook and HTC — while the software and workflow tooling sectors are already dominated by giants — AutoCAD, Unity, Amazon’s Lumberyard, and Epic Games with their Unreal engine — leaving limited opportunities for VC-backed plays.
But I believe significant opportunity exists in the content space, and in particular, supplying “picks and shovels” — or in this case polygons — to the rapidly growing demand for VR and 3D content.
As the VR industry scales, it must now power its own content revolution. This starts with the 3D models that act as the building blocks for any VR content. Let’s say that you want a can of Coke shown in virtual reality — someone needs to create a wireframe, which is then dressed in a pattern and rendered by the 3D engine. As adoption grows, we will see a spike in demand for more (and more complex) VR environments, and therefore more models.
A simple solution for bridging this gap between supply and demand are marketplaces for existing 3D models. Platforms such as CGTrader and TurboSquid today provide access to millions of modelled items, but these tend to be relatively simple graphics often referred to as ‘low polygon’ or ‘low poly’ models due to the lower complexity of the images.
These marketplaces follow the model of other platforms that exist for stock imagery, music, and design. Splice, for example, provides access to royalty-free samples for music production and it has spawned a thriving sub-industry of content creators specifically focused on Splice samples. It should, therefore, come as no surprise that TurboSquid was acquired earlier this year by Shutterstock, a leading platform for stock photography.
However, the dominance of low poly models on these marketplaces does not lend itself to complex or bespoke 3D graphics. And this has given rise to marketplaces for 3D modelling talent. CGHero, a Manchester-based start-up, enables production houses, studios, and corporates to access a vetted marketplace of 3D modelling talent that can scale flexibly with the demands of a project.
These talent marketplaces lend themselves to briefs that require more complexity. Unlike other talent markets, where individual contributors can be hired and slotted into a project relatively easily, creating complex 3D content is more akin to a shipbuilding, where orchestration is important and specialist craftspeople come at different times to do specialised jobs (from modelling and rigging, to hair and grooming). I do expect, however, that over time we’re likely to see the rise of automation in 3D content creation.
Computer vision has the capacity to convert simple 2D imagery into 3D models. Matterport, which recently went public at a $2.3bn valuation through a SPAC merger, has spent the past decade establishing itself as the leading provider of hardware and software for translating images into 3D models for the real estate industry.
And we’re now beginning to see academics and start-ups turn to broader uses of computer vision for 3D modelling. In 2020, researchers from Imperial and Facesoft.io, a London-based start-up using AI and machine learning for facial analysis, presented AvatarMe, a method for translating a single ‘in-the-wild’ image into a photorealistic 3D model. This 3D facial imagery could, for example, be used for VR meeting rooms.
Nonetheless, full-scale automation of 3D content creation is likely to be a way off, and there may be a need for solutions that reduce the hours of human input required to make 3D models. Promethean AI, founded in Los Angeles, harnesses AI to speed up the process of human artists creating artwork for video games. As reported by Venturebeat, an artist could simply ask the tool to ‘make a bedroom’ and it would generate a virtual bedroom that could be refined and customised by the artist.
Whether the approach lies in databases of content, human talent, or computer-driven solutions, we are likely to be an array of ‘picks and shovels’ to serve the growing audience of VR users worldwide.
