Why Video Post-Production and VFX will inevitably move to the cloud
The video post production and visual FX industries are massive users of compute and digital storage and they are therefore massive targets for Amazon, Microsoft Azure and Google Cloud. The prospect of exabytes of storage, hundreds of thousands of render hours and huge file transfers has all the public cloud providers collectively drooling, and for good reason - the professional video creation industries are some of the last to move to the cloud and come with the enormous marketing cache of "use the same cloud that XXX movie was made on" -Example.
The usual reasons you will hear for companies moving to the cloud - lower capital expenditure (CapEx), flexibility to spin up extra resources quickly and matching customer demand with compute resources are valid in post production and visual FX but I don't believe will be the key to cloud migration for the biggest video production users. I'm not going to argue that these economic factors are not important, they will be for some - especially smaller post/VFX houses that typically only handle 1 or 2 productions at a time and when not working need to 'hibernate' the company until the next big gig comes along. But for large production houses, those that can afford the equipment capex, can amortize those costs over many years and that have multiple projects running at a time - these purely economic calculations have proven insufficient to make the switch to the cloud thus far.
Instead I believe there are four factors, that are not commonly considered in this migration, but will actually will be the crucial ones to finally drive the video creation industry to the cloud:
Staff Flexibility, not Infrastructure Fexibility
The largest cost at major post production/VFX houses is not the cost of the equipment but the cost of recruiting, training and retaining quality digital artists. Really good artists are hard to find in what is now a global market and they can command high salaries. Paying these salaries, especially if the company does not have a full book of projects, can be prohibitive and can lead to low-balling projects to ensure that there is some work coming in to keep talent occupied during the slow periods. This short termism to keep staff working lowers pricing and margins across the industry.
The opportunity to scale the technology when projects come and go but also to scale the staff to match those projects is the true nirvana of the video creation industries. So where does the cloud come in? Once the content is in the cloud the workforce accessing that content can be global. The cloud also allows second by second secure access to content so a virtual pool of artists can be quickly spun up, assigned to a creative project and then removed from the project as soon as their specific task is done. It will be the true 'gig economy' but for high-end production talent where every artist is a freelancer and can be easily found and assigned to tasks. That dynamic talent model is only possible when the all the media required for a production is available in the cloud.
Some would say the cloud is insecure - it is after all the internet that we're trying to secure from this unreleased media. However, I would argue that the cloud is only as secure (or insecure) as any specific implementation of it. Put another way - the cloud is extremely secure as long as it's configured correctly. What is insecure is the current system of point to point transfers, multiple copies of files floating around between vendors, unencrypted hard drives in the back of a truck and relying on partners to delete files when they're supposed to.
A secure cloud implementation where files don't need to be moved to get from one production office to another, with accurate asset tracking, no duplicates, individualized logins and thorough audit trails is a compelling reason why productions should insist on the cloud - for their own protection.
We live in a world where all episodes of a show need to be delivered at once, where the premium delivery system is the internet itself and yet delivery of final assets often still requires multiple hard drives and a kid on a motorbike or point to point transfers using adaptations of IP protocols.
If all production occurred in the cloud, where all mezzanine files were assembled on secure cloud storage, then publishing those assets to distributors becomes as simple as pressing a big red "PUBLISH" button. The files need not move. The distributor is notified the finished content is ready for them, the access rights are changed and "delivery" just occurred instantaneously. When time is precious at the end of a production this can save hours or even days of copying files between drives and locations. Increasingly we'll need all those days to be able to continue to refine the quality of the production.
High end VFX are traditionally rendered on huge CPU based render farms - and the good thing about all that CPU power is you can apply the same processors for all sorts of other intensive compute tasks. Now that we're seeing true ray tracing on GPUs this game is going to change but it will not be economical for every major post house to create a new dedicated GPU farm, just for ray tracing.
Enter the big cloud providers who have the capital and scale to be able to offer the latest GPUs, when they are released at massive scale and in multiple regions. This will allow the VFX and post-houses to grow into GPU computation and use that power for specific tasks. With the speed of innovation in ray tracing specific silicon only likely to accelerate, it will be best to leave the cloud companies to keep making those investments as this #Turin based refresh cycle is only just beginning.
Look for the cloud migration to pick up speed in 2019 - and the reasons listed above will tip the economics so that this move will be a fast one. Those left behind will suffer with increasingly inefficient workflows, the inability to compete on all projects and from 2020 they'll start to see productions require cloud production systems from all vendors…