The paper addresses the gap between high-end and low-end methods in 3D face modeling. It introduces FLAME (Faces Learned with an Articulated Model and Expressions), a model designed to work with existing graphics software and be easy to fit to data. FLAME is trained from over 33,000 3D scans and combines a linear shape space with articulated jaw, neck, and eyeballs, pose-dependent corrective blendshapes, and additional global expression blendshapes. The model is more accurate and expressive than existing models like FaceWarehouse and Basel Face Model. The paper also describes the registration process, which involves aligning a template mesh to scan sequences and making the registrations available for research purposes. The registration process is fully automatic and includes co-registration and image texture to achieve high-quality alignment. The paper compares FLAME to other models by fitting them to static 3D scans and 4D sequences using the same optimization method. FLAME is significantly more accurate and is available for research purposes.The paper addresses the gap between high-end and low-end methods in 3D face modeling. It introduces FLAME (Faces Learned with an Articulated Model and Expressions), a model designed to work with existing graphics software and be easy to fit to data. FLAME is trained from over 33,000 3D scans and combines a linear shape space with articulated jaw, neck, and eyeballs, pose-dependent corrective blendshapes, and additional global expression blendshapes. The model is more accurate and expressive than existing models like FaceWarehouse and Basel Face Model. The paper also describes the registration process, which involves aligning a template mesh to scan sequences and making the registrations available for research purposes. The registration process is fully automatic and includes co-registration and image texture to achieve high-quality alignment. The paper compares FLAME to other models by fitting them to static 3D scans and 4D sequences using the same optimization method. FLAME is significantly more accurate and is available for research purposes.