While audio recordings and observation might have dominated past decades of mathematics education research, video data obtained from classrooms or design experiments is now the dominant form of data in the field. Alongside this development has been an increased interest in the role of the body in teaching and learning mathematics. In this paper, I argue that video research in mathematics education must be situated within the history of scientific cinema, and studied for how it produces a particular image of the human body. I compare and contrast current video research with early cinematic attempts to capture the gestural sensory-motor movements of bodies, and discuss the links between this kind of research and the shifting structures of industrial and digital labour. I unpack how video practices are now transposed through new algorithmic analytics of large video data sets, within a global software culture of surveillance that extracts value from human bodies. My argument uses theoretical tools from Gilles Deleuze - in particular his ideas on cinema and the moving image - to show how video data might be reconceived.