Faceware’s new pricing and deep learning

Faceware specializes in face tracking and solving, both hardware and software. Faceware will be at SIGGRAPH 2019 in LA with some amazing new demos but we caught up with them prior to the show to discuss the changes the company has been implementing for the last 3 months. The company has recently adjusted its pricing and moved into providing more accurate solutions with Deep Learning.

Faceware Technologies was established in 2012 after years as part of the facial tracking and simulation company Image Metrics. It aims to service the needs of professional animators in the game, film, television, and entertainment industries. The Faceware Facial Motion Capture product line has been utilized in the production of hundreds of titles and films and it is a facial solution provider for clients such as DNEG, Digital Domain, Blur Studios, Activision-Blizzard, Rockstar Games, Microsoft, Electronic Arts, Ubisoft, Sega, Sony, Motion Theory and MPC.

New Pricing

Faceware provides a range of software and hardware solutions for facial capture.

Faceware’s product consists of the Faceware Indie and MkIII Headcam Capture Systems;

  • Faceware Analyzer, which allows clients to analyze and process their own performance videos;
  • Faceware Retargeter, an Autodesk plugin which allows users to create facial motion capture data at a much faster rate than traditional methods; and
  • Faceware Live, the real-time facial capture and animation solution.

There is also a live client for UE4 made by Glassbox.

In line with similar moves from Adobe, Autodesk, Maxon, and others, the biggest change was to offer their tools as an annual base license, starting at just $99 a year.

They did this with a new logo and look, plus a new site, including an online store. “Previously if customers that wanted to buy licenses would send us an inquiry, we’d send them a quote and there was kind of this back and forth that would take sometimes two, three days. Now there’ll be able to go directly on our store, buy products and have licenses in their hands, the same day” commented Peter Busch VP of Business Development.

You can still buy the full license, the new lower cost perpetual licences are for Analyzer and Retargeter now start at $750 and $500 respectively, including the first year of support.

They also have new educational pricing. The Academic pricing now starts at $50 for some software licenses, as well as $2,500 for complete turnkey systems (hardware and software) for academic customers, including the first year of support. “We now have a full-time academic business development manager on the team and with his help, we’ve crafted a new program that aims to grow the number of schools using our products,” commented Busch. “We’ve got about a hundred schools now worldwide, with our tools in their curriculums, but we want to grow that more.”

Deep Learning

“Our Maya and 3Dstudio Max and MotionBuilder plugins are now integrated with about two years worth of deep learning research, that leveraged our proprietary Dataset to specifically make some animation improvements”, says Busch.

The last releases of Analyzer 3.8 and Retargeter 5.8 integrated almost 2 years of research and development utilizing Deep Learning Techniques. It used a proprietary data set of specific facial capture images. Their research annotated and used a targeted subset of synthesized data of nearly 3 million images to re-build from the ground-up their approach to accurate animation of the jaw. This deep learning technology poses no changes to the workflow and is deployed through Faceware’s normal Retargeter’s AutoSolve feature, giving users a better starting point from which to start animating a facial performance. Jaw positioning and animation is critical to lip sync, one of the hardest things to get correct in facial capture.

Faceware collected about three quarters of a million images specific to facial capture. They dont use every frame in an image sequence, “we only need to collect key expressions from a shot” Busch explains. Not all of Faceware’s users choose to opt-in, but enough did and they submit images over an 18 month period. “That’s representative of tens of thousands of different subjects, ” he adds. “And then within each subject you’ve probably got anywhere from 20 to 200 different expressions of those subjects”. As deep learning requires both a lot of training data and a wide variety, or breadth of examples, the company was able to get great results on their jaw estimator.

Shepherd

Faceware also launched late last year in closed Beta their Shepherd motion capture sync software. This software automates Faceware’s Mk III headcam recording, and integrates with industry body motion capture systems, such as Vicon and Optitrack, to sync facial and body mocap recordings on Windows PCs and tablets.

“In most mocap recording sessions, operators are working with one software package for body mocap, and another for facial mocap. This adds a lot of complexity and a lot of work for your operators,” said Peter Busch. “Shepherd syncs a studio’s face and body mocap pipelines, providing a more unified, succinct and automated motion capture process — something that is a key initiative for our users.”

Shepherd gives facial motion capture operators the ability to control recording and playback of multiple video recording devices. Shepherd simplifies motion capture shoots by eliminating the need to manually track take numbers, file names, and recording times. Because of Shepherd’s ability to start and stop facial recording at the body system’s direction, the facial motion capture operator is better able to focus on the actor’s performance, rather than managing the tech. Shepherd currently connects to Vicon and Optitrack mocap systems (with more integrations to come) to provide facial motion capture recordings that are in sync with body motion capture recordings.

Shephard moved from Beta in March and can now be purchased for US$499 / year.

Digic Pictures: LOVE DEATH + ROBOTS

Digic pictures were responsible for episode 16 of the Netflix series Love, Death + Robots. Busch points to the company as a key customer and indicative of the changed state of the industry. “Episode 18, the Secret War was produced by Digic Pictures in Hungry. Obviously, they have a great name for the incredible level of art they’ve done in the game cinematics and in the world of game trailers,” he comments. “And now they were entirely responsible for that episode, and you could really see the quality of their work. But it also speaks to a trend of studios that were building a name for themselves in one vertical and now expanding into streaming content. Animated streaming content is a very large vertical that continues to grow as these platforms grow and we are seeing more and more studios needing to approach work like Digic are now producing”.