Post Production in the Cloud
By Adrian Pennington, Post Perspective
Talked about for years, the capacity to use the cloud for the full arsenal of post workflows is possible today with huge ramifications for the facilities business.
Rendering frames for visual effects requires an extraordinary amount of compute power for which VFX studios have historically assigned whole rooms full of servers to act as their renderfarm. As visual quality has escalated, most vendors have either had to limit the scope of their projects or buy or rent new machines on premises to cope with the extra rendering needed. In recent times this has been upended as cloud networking has enabled VFX shops to relieve internal bottlenecks to scale, and then contract, at will.
The cloud rendering process has become so established that even this once groundbreaking capability has evolved to encompass a whole host of post production workflows from previz to transcoding. In doing so, the conventional business model for post is being uprooted and reimagined.
“Early on, global facility powerhouses first recognized how access to unlimited compute and remote storage could empower the creative process to reach new heights,” explains Chuck Parker, CEO of Sohonet. “Despite spending millions of dollars on hardware, the demands of working on multiple, increasingly complex projects, simultaneously, combined with decreasing timeframes stretched on-premise facilities to their limits.”
Above: Chuck Parker, Sohonet CEO
Public cloud providers (Amazon Web Services, Google Cloud Platform, Microsoft Azure) changed the game by solving space, time and capacity problems for resource-intensive tasks. “Sohonet Fastlane and Google Compute Engine, for example, enabled MPC to complete The Jungle Bookon time and to Oscar-winning standards, thanks to being able to run millions of Core hours in the cloud,” notes Parker.
Small to mid-sized companies followed suit. “They lacked the financial resource and the physical space of larger competitors, and initially found themselves priced out of major studio projects,” says Parker. “But by accessing renderfarms in the cloud they can eliminate the cost and logistics of installing and configuring physical machines. Flexible pricing and the option of preemptible instances mean only paying for the compute power used, further minimizing costs and expanding the scope of possible projects.”
Milk VFX did just this for rendering its Oscar-winning sequences on Ex Machina. Without the extra horsepower, the London-based house could not have bid on the project in the first place.
“The technology has now evolved to a point where any filmmaker with any VFX project or theatrical, TV or spot editorial can call on the cloud to operate at scale when needed — and still stay affordable,” says Parker. “Long anticipated and theorized, the ability to collaborate in realtime with teams in multiple geographic locations is a reality that is altering the post production landscape for enterprises of all sizes.”
Parker says the new post model might look like this. He uses the example of a company headquartered in Berlin — “an innovative company might employ only a dozen managers and project supervisors on its books. They can bid with confidence on jobs of any scale and any timeframe knowing that they can readily rent physical space in any location, anywhere in the world, to flexibly take advantage of tax breaks and populate it with freelance artists: 100 one week, say, 200 in week three, 300 in week five. The only hardware (rental) costs would be thin-client workstations and Wacom tablets, plus software licences for 3D, roto, compositing and other key tools. With the job complete, the whole infrastructure can be smoothly scaled back.”
The compute costs of spinning up cloud processing and storage can be modelled into client pitches. “But building out and managing such connectivity independently may still require considerable CAPEX — one that might be cost-prohibitive if you only need the infrastructure for short periods,” notes Parker. “Cloud compute resources are perfect for spikes in workload but, in between those spikes, paying for bandwidth you don’t need will hurt the bottom line.
Dedicated, “burstable” connectivity speeds of 100Mbit/s up to 50Gbit/s with flexibility, security and reliability are highly desirable attributes for the creative workflow. Price points, as ever, are a motivating concern. Parker’s offerings “move your data away from Internet bandwidth, removing network congestion and decreasing the time it takes to transfer your data. With a direct link to the major cloud provider of your choice, customers can be in control of how their data is routed, leading to a more consistent network experience.
“Direct links into major studios like Pinewood UK open up realtime on-set CGI rendering with live-action photography for virtual production scenarios,” adds Parker. “It is vital that your data transits straight to the cloud and never touches the Internet.”
With file sizes set to continue to increase exponentially over the next few years as 4K and HDR become standard and new immersive media like VR emerges to the mainstream, leveraging the cloud will not only be routine for the highest budget projects and largest vendors, it will become the new post production paradigm. In the cloud creative workflows are demystified and democratized.