Performer2 Multi-View software, a true game-changer in facial motion capture at SIGGRAPH 2018
Cesson-Sévigné, France – July 26, 2018.
Dynamixyz has released Performer2 Multi-View, the world’s first commercially-available software solution to run markerless facial capture on multi-camera streams, in-house, unlimited. Come experience our live demo and treat yourself with the ultimate markerless facial motion capture experience.
Tailored for demanding productions, Performer2 Multi-View breaks new grounds for markerless facial motion capture with unprecedented 3D precision and organic-looking facial animations, that will definitely boost your productions quality. Dynamixyz Performer2 Multi-View makes the most of multiple camera angles, improving both robustness and precision thanks to an accurate 3D acquisition of the actor’s performance.
Performer2 Multi-View is, for sure, a productivity-enhancing technology as less time is spent on tedious steps (annotations, occlusions management…) while still benefiting from Performer’s efficient workflow (first-rate raw results, batch mode processing, fully-automatic production once setup is achieved…).
SIGGRAPH 2018, which is being held in Vancouver from August 12th to 16th, will be an opportunity to present this brand new technology. Dynamixyz will perform a live demonstration of Multi-View showcasing a Chest Mounted stereo Camera system.
This is a great chance to discover real-time and to discover with what precision the actor’s expressions are faithfully transferred to the 3D character giving lifelikeness and aesthetic results.
Some technical insights
Two technologies are co-existing in the facial motion capture market: markerless and marker-based. Dynamixyz is one of the few players providing markerless solutions, although the system is also able to process marker tracking.
Marker-based motion capture makes sense for body tracking: a body is well-approximated as an articulated ensemble of rigid limbs, so a sparse set of well-located measurements -the markers- is sufficient. A face however, expresses meaning and emotions through the motion and deformation of facial skin, exhibiting complex, continuous patterns of appearance changes. It is thus best captured and analyzed through a dense analysis of the face geometry and texture, where every pixel acts as a measurement. That is the purpose of markerless technologies.
Whether in the context of live event or just-in-time virtual production (previz), real-time tracking and retargeting are highly-anticipated capabilities. With a dedicated pipeline taking advantage of training data, Dynamixyz proves to be part of the leading references.
Hardware requirements and recommendations
• All cameras must be gen locked, and have the same resolution.
• Supports up to 9 cameras (though optimized setups are 2, 3 or 4 cameras).
• Camera layout is free (supports vertical, horizontal, square layouts).
• Minimum distance between camera and face: 150mm, with a recommended distance between two cameras starting from 80mm.
• Performer2 Multi-View is compatible with any multi-camera head gear.
• Free-head mode available (regular static cameras) with the same above-mentioned requirements.