DI4D, a specialist in performance-driven facial animation services, has launched a new service called PURE4D, a precise pipeline for digital double facial animation.
PURE4D combines their existing DI4D PRO service and head-mounted camera (HMC) service with machine learning to produce scalable, high-fidelity facial animation for video games. It captures an actor’s facial performance as it happens, without relying on artist corrections or enhancements, delivering realistic facial animation ready to use in game engines.
The DI4D PRO system acquires high fidelity 4D facial expression data, and the DI4D HMC system captures accurate 4D facial performance data, simultaneously with body motion capture. Machine learning algorithms efficiently combine both sets of data to increase the scalability of the system and make it useful for large projects.
Producers of video game projects and in-game animation now have the opportunity to match the accuracy and realism that previously required complex VFX pipelines, and had to be reserved for feature film and pre-rendered cinematics. Moreover, this new process doesn’t need a traditional facial animation rig as part of the capture process, which helps control the complexity and cost of facial mocap.
DI4D’s speciality is its proprietary 4D reconstruction pipeline and services developed for performance-driven facial animation. It precisely reproduces minute details of an actor’s facial performance, and reconstructs a 3D scan per video frame. The detailed surface information serves as a foundation used to generate animations that preserve the subtle expressions of an actor’s performance, resulting in believable digi-doubles and virtual human characters.
The company’s main service, DI4D PRO, captures sequences of high resolution 3D facial performance data, in colour, using nine synchronised 12-megapixel machine vision cameras. The process produces photoreal data closeup shots, for example, or can record the exact movements of facial performance used to create extreme high fidelity FACS, ROMS and visemes for rigging and machine learning.
The HMC service suits projects that need facial animation in scenes with dynamic, spontaneous performances. The light head-mounted rig is fitted with two small, synchronised high-res video cameras, on-board LED lights and a compact video recording system. Together the components work as a stereo photogrammetry system, recording at a resolution of 3mp at 60 fps, and can capture facial-performance data simultaneously from multiple actors, along with a full-body motion capture system.
DI4D’s optical flow tracking software accounts for every pixel and feature from each image, delivering direct point-cache animation applied to the mesh. Every vertex in the mesh is tracked for consistency and detail. di4d.com