Imagine you’re working on a VFX-intensive show that calls for a lot of Houdini. Before you begin, you consult with Athera.
Within minutes, you learn how much your project will cost, what types of artists are qualified to work on it, and what equipment you’ll need.
A cost prediction tool like the one imagined above is just one of many applications the Athera team wants to bring to the industry in the not too distant future – with cloud as the necessary ingredient.
Today when we talk about cloud, we tend to think of it in terms of the flexible option it provides for tackling tasks that can also be done without it.
But what about the things we might be able to do in the industry only because of cloud? Things like the cost prediction wizard imagined above?
Armed with the compute power of cloud, the Athera team has been exploring some game-changing ideas that could be realized using technologies like machine learning, AI, and blockchain – ideas that may be put into practice sooner than you think.
1. Asset tagging
We all know that for every scene that makes it to the screen, there are hundreds of hours of footage left on the cutting room floor. Yet every single item in any one of those frames continues to be an asset that could potentially lend itself well to your next project – saving time and money.
But how can you easily find what you’re looking for amid the masses of media you have at your disposal?
The solution is categorization.
A lot of the bigger studios began to recognize the economic advantage they could gain by tagging and tracking assets and re-using them.
But studio categorization practices of today remain inefficient, lacking a common vocabulary for artists to work with. What one artist tags a ‘butterfly,’ another might tag a ‘moth,’ while still another might call it a ‘flying insect.’
“The moment you get a human involved, it’s going to be inconsistent, error prone, and not scalable,” says Mathieu Mazerolle, Athera senior product manager.
Categorize better with machine learning
So how do you make your categorization consistent, error free, and instantaneous? The answer is single image video analysis using machine learning.
To work well, machine learning just needs a pool of data – just like the huge amounts of media available at a studio’s disposal – and a correlation algorithm.
“You can pretty well have millions and millions of frames of data out there and it seems daunting, but if you can throw millions and millions of computers looking at each frame, it goes really quickly,” says Mazerolle, pointing to cloud’s role in the process.
A cloud-based tagging system might allow you to do other interesting things as well – like intersecting cost analysis with image categorization to provide you with valuable information, such as identifying which elements exist in your priciest shots.
“This idea of dimensionalizing the data set and tagging it, and reasoning about it, and figuring out what aspects of those things are human relatable is something that the cloud does really well,” explains Mazerolle.
2. Cost prediction and rendering workflows
The Athera team is also working on the idea of a cost prediction tool that could transform project management in the industry.
The idea is that if everybody’s projects and budgets were aggregated anonymously and you had a machine learning tool looking at trends within that giant pool of data, that same tool could make all kinds of useful suggestions to you – such as what your budget would look like on a job, or what kind of people would be qualified to work on it.
“We think what’s really going to change the game through cloud is what happens when lots of data become aggregated and we can start making interesting suggestions to people based on a bunch of data that’s not related directly to what they have, but related to how similar their projects are to all the other projects in the cloud,” says Mazerolle.
Better prediction on your rendering workflows
Machine learning derived cost prediction could become especially useful applied to rendering workflows – normally difficult to set time and cost budgets to with each job so unique in complexity.
Assisted by machine learning, Athera could simply look at what you’re trying to render and ask you when you want it, and then tell you – with great accuracy – what it would cost if you want it in a week, today, or within the hour.
“Instead of trying to figure out all of these complicated things like how much time is this frame going to take, and how many frames do I have, and should I split it up – the system just speaks plain language with you about how fast you want it and gives you the cost, and from there you can make informed decisions about trading off time and money,” says Mazerolle.
Machine learning can make such time-cost inferences by looking at hundreds of thousands of jobs that have characteristics that are similar to the one you’re proposing, and knowing what the time and cost tradeoffs of those were.
The only thing needed to make that work is a good algorithm and lots of data points to feed it – something the Athera team expects to have in the not too distant future as more and more people move to cloud.
3. Reducing drudgery in applications:
The Athera team hopes cloud may also one day wipe out some of the more manual processes associated with the industry’s more time-consuming jobs – like rotoscoping.
The more elements you have and the more elements you’re combining, the more the work becomes complicated and the more people you need, because the only way to do more roto work is to throw more humans at it, right?
Maybe not, says Mazerolle, if Athera can start using cloud to do it instead.
“If you could take all of the roto work that was done on all the projects in the world and you ran a machine learning algorithm to figure out what did the humans do to get a quality shot – like how did they handle the hair – you could actually do a pretty good job,” he says, explaining this particular algorithm approach would be example based, using human behaviour to synthesize a new computer driven behaviour.
The Athera team believes cloud is the key to removing the drudgery from many other compute-intensive tasks in the industry too – something like de-noising, for example, where you would want every single part of your workload potentially backed by thousands of processors, not just the one that’s on your desktop.
“So generally where you’re either doing a lot of manual work that’s similar to a lot of other manual stuff and machine learning could just do it – or things where there’s a lot of compute power, and you’re just going to be in bottleneck by that – cloud can actually solve both those problems to make it so the human doing the work can actually focus on the right things,” says Mazerolle.
4. Job brokering between artists using blockchain
If you’ve heard of blockchain, you’re likely thinking of its connection to virtual currencies.
But the Athera team has other things in mind for the technology, looking to the very practical applications it could have for building contracts in the industry.
Blockchain is simply a growing list of records, called blocks, that are linked together through cryptographic signature.
The technology allows you to be absolutely certain that the contents of each record have not been tampered with. You also have proof that only a certain person could have created that record.
Using blockchain in the world of VFX, it would be conceivable to have a completely digital, completely cloud based way of asserting that work is done and shipped between clients and artists. You can validate it, and it becomes a contract that anyone in the world can look at.
“It would allow you to do things like shoot stuff off to freelancers without even knowing who they are, which would no longer be important, as long as the work got delivered and you signed off onto it and it was signed off through something reliable that everybody can assert against, like blockchain,” says Mazerolle.
While contracts can already be sent back and forth digitally, it’s all done manually and becomes complicated by the fact that everyone does it differently.
“Everyone’s got a different bank account, some people want credit card, some people want cheque, people go out of business, people deliver the goods but it turns out it’s not really what you wanted so you don’t pay the guy, then he disputes it,” all leading to a bit of a mess, explains Mazerolle.
Instead, he says, if everything was purely digital from the get go, it would make sense to move the money digitally as well, and have the movement of that money contingent on something that’s digital too – like blockchain.
“Our idea with blockchain is making it more about being the broker of the work and the payment and not just the broker of the infrastructure of what people use to do the work,” says Mazerolle.
It’s a big idea for the future, he says, which his team believes will be a natural extension of using Athera to do VFX work.
“We really think these technologies like machine learning and blockchain are revolutionizing a lot of things and we’re really excited to apply them to Athera because we’re really thinking this is the opportunity that lies beyond moving people to cloud – this is the opportunity of what new things happen when we’re on the cloud,” he says.
Are you ready for the cloud to supercharge your next project? Contact us about how we can get you started!