These may sound like physical human ailments but for Framestore, the pain is virtual. Or rather, it was. With the introduction of IKinema Action for MotionBuilder, the Framestore Capture Lab team is transforming the way they transfer motion capture data to animated characters. Technical Supervisor Gary Marshall and Studio Manager Richard Graham are in the middle of a large project and realism is key.
With a real-time solve pipeline based on Vicon Blade and Autodesk MotionBuilder, their existing workflow is ideal for previs work, where the priority is to deliver data without the time overhead of traditional mocap post-processing. For the last couple of years, most of the projects they’ve been working on are just that – previs shoots for Framestore partner The Third Floor.
However, when they started work on what is quite an unusual project, it became apparent the workflow wasn’t delivering the necessary realism. Marshall is careful not to reveal too many details – the project is under tight wraps until its launch in October – but explains in broad terms. "We did a number of shoots using the old pipeline and ran into challenges around the fidelity of the data," he says. "We went backwards and forwards with the Shot Supervisor and Crowd Supervisor – of course there’s always a certain amount of cleaning up required – but because they get the data so quickly, they’re used to this pipeline. It was only after they’d taken it through the crowd engine and were seeing what it looked like in final shots that they came back to us to say ‘we don’t really like the shoulder and spine movement'."
IKinema Action for MotionBuilder
Framestore prides itself on being the best in the business when it comes to producing realistic animation so the team needed to find a solution – and one that fit within their existing pipeline. Marshall knew IKinema had a growing reputation for high quality solving, so when he met IKinema founder Alexandre Pechev at SIGGRAPH last year, he discussed the challenge with him. "Alex saw some of those issues with spine/shoulder solving and said 'this is screaming out for an IKinema plugin for MotionBuilder'," says Marshall. "I said I’d definitely be up for testing it and helping you guys develop it." Fast forward a few months and Action for MotionBuilder was added to the pipeline.
"The results with IKinema Action for MotionBuilder were so much better, we had no choice really. Even though we have a mass of data to reprocess, it will be worth it."
"We did an entire day of motion capture then looked at the data – instantly the feedback was really great. For the supervisors, it made a world of difference – the shoulders, the spines, the motion in general was looking more natural - validating the reason to use mocap in the first place," he explains.
In fact, feedback has been so good that the Capture Lab team has been asked to go back through all the raw data and reprocess it through IKinema Action. "In the end, all the crowd motion for this project will have gone through this process," says Marshall, pointing out that after three and a half months in production, that amounts to a lot of data. "Shooting wasn’t every day, about one day a week, but it’s still quite a volume of work," he says. "Big data is a huge part of this story."
Framestore CG Supervisor Alan Woods agrees: The results with IKinema Action for MotionBuilder were so much better, we had no choice really. Even though we have a mass of data to reprocess, it will be worth it.”
One of the advantages of working with IKinema is flexibility: the IKinema solver is the same, in terms of animation quality and setup, for post production and live shoots. Framestore currently plans to use Action for MotionBuilder for all their non-previs work – and sees potential beyond that. "It’s been a good partnership, working with the IKinema team," says Marshall. "They’re interested in helping us develop tools around VR and real time, as well as how we could potentially leverage their technology in our on-set endeavours – deriving motion from reference video.”"
The partnership was established several years ago. The Capture Lab team was experimenting with their solving workflow for Wrath of the Titans and took Pechev’s advice to do the heavy lifting with their IKinema Action for Maya solver. It worked very well, says Marshall. However, the difference between Wrath of the Titans and the current project is they’re now able to work within MotionBuilder, which is more efficient he explains: "Because MotionBuilder is a real-time engine, once you have your solve going – on a template – you can then import your motions in from, say, your morning sessions, preview them instantly and scrub them back and forth, whereas in Maya you never could see it in real-time – you had to setup your solve, and then render it. And if you made a tweak to a parameter in the solve, you’d then have to render it out again." Now, with Action for MotionBuilder, they can even work live if they want to.
However, Marshall stresses the key benefit of working with Action for MotionBuilder is more about quality than speed: "Realism is key to believable characters and the new workflow delivers that realism by retaining the fidelity of the original performance. With IKinema Action you are using the raw marker data that’s coming out of Blade and you’re using that to drive your character, rather than starting from a presolved skeleton and retargeting that solve. For us, it’s not even about a speed-up – it’s more about optimization."
The Framestore team knows that as customer expectations for animation keep rising, it needs to continually raise the bar. Action for MotionBuilder was the ideal solution: high quality rapid solving that fits right into their existing pipeline. A perfect solution delivering state of the art realism.
Deliver stunning animation from quality MoCap
Action for Maya and Action for MotionBuilder incorporates IKinema’s core rigging technology to solve motion capture data delivering stunning animation every time. Match live actor performance to virtual avatars with greater realism, and eliminate post process work by automatically resolving inaccurate data, gaps and noise. Fluid spine, neck and shoulders are easily achieved while maintaining solid lock to floor.
See more articles
Please, introduce yourself: