Tag Archives: Foundry Katana

VFX pipeline trends for 2020

By Simon Robinson

A new year, more trends — some burgeoning, and others that have been dominating industry discussions for a while. Underpinning each is the common sentiment that 2020 seems especially geared toward streamlining artist workflows, more so than ever before.

There’s an increasing push for efficiency; not just through hardware but through better business practices and solutions to throughput problems.

Exciting times lie ahead for artists and studios everywhere. I believe the trends below form the pillars of this key industry mission for 2020.

Machine Learning Will Make Better, Faster Artists
Machines are getting smarter. AI software is becoming more universally applied in the VFX industry, and with this comes benefits and implications for artist workflows.

As adoption of machine learning increases, the core challenge for 2020 lies in artist direction and participation, especially since the M.O. of machine learning is its ability to solve entire problems on its own.

The issue is this: if you rely on something 99.9% of the time, what happens if it fails in that extra 0.1%? Can you fix it? While ML means less room for human error, will people have the skills to fix something gone wrong if they don’t need them anymore?

So this issue necessitates building a bridge between artist and algorithm. ML can do the hard work, giving artists the time to get creative and perfect their craft in the final stages.

Gemini Man

We’ve seen this pay off in the face of accessible and inexpensive deepfake technology giving rise to “quick and easy” deepfakes, which rely entirely on ML. In contrast to these, bridging the uncanny valley remains in the realm of highly-skilled artists, requiring thought, artistry and care to produce something that tricks the human eye. Weta Digital’s work on Gemini Man is a prime example.

As massive projects like these continue to emerge, studios strive for efficiency and being able to produce at scale. Since ML and AI are all about data, the manipulation of both can unlock endless potential for the speed and scale at which artists can operate.

Foundry’s own efforts in this regard revolve around improving the persistence and availability of captured data. We’re figuring out how to deliver data in a more sensible way downstream, from initial capture to timestamping and synchronization, and then final arrangement in an easy, accessible format.

Underpinning our research into this is Universal Scene Description (USD), which you’ve probably heard about…

USD Becomes Uniform
Despite having a legacy and prominence from its development with Pixar, the still relevant open-sourcing and gradual adoption of Universal Scene Description means that it’s still maturing for wider pipelines and workflows.

New iterations of USD are now being released at a three month cadence, where before it used to be every two months. With each new release comes improvements as growing pains and teething issues are ironed out, and the slower pace provides some respite for artists who rely on specific versions of USD.

But challenges still exist, namely mismatched USD pipelines, and scattered documentation which means that solutions to these can’t be easily found. Currently, no one is officially rubber stamping USD best practice.

Capturing volumetric datasets for future testing.

To solve this issue, the industry needs a universal application of USD so it can exist in pipelines as an application-standard plugin to prevent an explosion of multiple variants of USD, which may cause further confusion.

If this comes off, documentation could be made uniform, and information could be shared across software, teams and studios with even more ease and efficiency.

It’ll make Foundry’s life easier, too. USD is vital to us to power interoperability in our products, allowing clients to extend their software capabilities on top of what we do ourselves.

At Foundry, our lighting tool, Katana, uses USD Hydra tech as the basis for much improved viewer experiences. Most recently, its Advanced Viewport Technology aims at delivering a consistent visual experience across software.

This wouldn’t be possible without USD. Even in its current state, the benefits are tangible, and its core principles — flexibility, modularity, interoperability  — underpin 2020’s next big trends.

Artist Pipelines Will Look More Iterative 
The industry is asking, “How can you be more iterative through everything?” Calls for this will only grow louder as we move into next year.

There’s an increasing push for efficiency as the common sentiment prevails: too much work, not enough people to do it. While maximizing hardware usage might seem like a go-to solution to this, the actual answer lies in solving throughput problems by improving workflows and facilitating sharing between studios and artists.

Increasingly, VFX pipelines don’t work well as a waterfall structure anymore, where each stage is done, dusted, and passed onto the next department in a structured, rigid process.

Instead, artists are thinking about how data persists throughout their pipeline and how to make use of it in a smart way. The main aim is to iterate on everything simultaneously for a more fluid, consistent experience across teams and studios.

USD helps tremendously here, since it captures all of the data layers and iterations in one. Artists can go to any one point in their pipeline, change different aspects of it, and it’s all maintained in one neat “chunk.” No waterfalls here.

Compositing in particular benefits from this new style of working. Being able to easily review in context lends an immense amount of efficiency and creativity to artists working in post production.

That’s Just the Beginning
Other drivers for artist efficiency that may gain traction in 2020 include: working across multiple shots (currently featured in Nuke Studio), process automation, and volumetric-style workflows to let artists work with 3D representations featuring depth and volume.

The bottom line is that 2020 looks to be the year of the artist — and we can’t wait.


Simon Robinson is the co-founder and chief scientist at Foundry.