In visual effects it is usually required to process multiple film plates in order to reproduce the performance of real actors using a digital 3D character. One of the most challenging aspects of this process is the complexity of facial animation. Motivated by the work required to capture and digitally augment the performances for the 2019 film Cats [Cats 2019], we created FaceLab, a system that drives the animation of a 3D facial rig by optimizing its blendshape targets' weights and the head transformation matrix to match the reference plate. The system is designed to scale across different scenarios with varying complexity and produced tangible savings on the show. CCS CONCEPTS • Applied computing → Media arts; • Computing methodologies → Motion capture; Machine learning.