One Advantage of Moving VFX Pipelines to the Cloud Is Making 360 Video Environments Much More Realistic
Foundry plans a summer launch of a beta version of Elara, a new cloud-based platform for post-production that combines storage infrastructure with creative applications like Nuke, V-Ray and Houdini.
Users would set up a “virtual studio” on the service, where their data would live alongside required creative tools and a dynamic render farm. They would access those tools via a simple web browser — any web browser. Because the software, project data, and computing power all reside in the cloud, only a stream of pixels depicting the user interface of a given tool would be delivered to the user.
At an NAB press briefing introducing Elara, Foundry CTO Jon Wadelton said Elara had its genesis three years ago as part of a government-funded project with ILM in the U.K. “How can we think about a world where a pipeline — you don’t have to invest in up-front hardware, you don’t have to do capacity planning, [and] you can think of this flexible resource being the cloud,” he said. “How can we work in that environment?”
During a demo, Foundry’s lead engineer on pipeline in the cloud, Simon Pickles, emphasized the service’s ease of use and flexibility, promising that APIs will allow Elara to be integrated into existing pipelines, including the use of data hosted on Elara with applications running locally. He also said the model was inherently secure. “There’s no longer any need to share sensitive project data with a remote collaborator,” he said. “By running the applications alongside the data, all your collaborators ever receive is a pixel stream.”
Elara also provides detailed analytics on a per-project and per-user basis, and Foundry’s co-founder and chief scientist, Simon Robinson, told StudioDaily that production tracking will likely be incorporated in the future. Robinson said there is no reason why software from other providers (obviously, we’re thinking Autodesk) couldn’t be provided through the service, though no deals are in place. And he said pricing terms have not yet been set.
Elsewhere at The Foundry
On the eve of NAB, Foundry made Nuke, NukeX and Nuke Studio 11.0 available as a beta for current Nuke customers with valid commercial licenses. Robinson said development of Nuke is focused on image processing and efficiency of workflow, including giving multiple artists the ability to collaborate at the same time on the same comp.
And Cara VR, a 360/VR plug-in for Nuke, is going strong as a high-end tool for VR and 360 content creation. “For us, it’s a bellwether of people’s growth of interest in VR, specifically 360 video,” Robinson said. “It’s still hard for us to work out how people will monetize it, but it’s going well on our end.”
The Importance of Positional VR
New developments in Cara VR could help address what’s commonly referred to as “six degrees of freedom,” or 6DoF, which Robinson calls “positional VR.” He’s referring to the ability of 360 environments to respond correctly to small changes in position of the viewer, introducing appropriate parallax effects to send a clearer signal to the eyes that they exist in real space. To date, those effects have mainly been limited to computationally generated 3D environments that can respond in real time to head movement rather than 360 video, which is locked to a single point of view at any given moment in time.
But Robinson said it will be possible to generate those effects in 360 video, as long as it’s possible to gather information about the position of objects in space, whether it comes from the current generation of 360 camera rigs that provide depth metadata, or from full-on light-field cameras in the future.
That makes sense — Lytro Cinema is already on board as an early Elara user. Since light-field workflow is so computationally heavy, Lytro’s plan has always been to do the processing in the cloud. And that ties into both Elara and the vision of positional VR.
“We see a future where compositing isn’t about compositing rectilinear images anymore,” Robinson said. “It’s about compositing dense volumetric data instead.”