This paper discusses theoretical and methodological issues arising from a video-based research design and the emergent tool "Joint Screen'"when grasping joint activity. We share our reflections regarding the combined reading of four synchronised camera perspectives combined in one screen. By these means we reconstruct and analyse multimodal moment-to-moment interaction between young peers engaged in an open-ended baking activity. We rely on a fine-grained analysis of three multimodally transcribed video extracts to highlight how a combined viewing of multiple joint camera perspectives provides access to how participants "do" joint activity through the simultaneous and continuous use of embodied resources. We argue that 'Joint Screen' generates an 'expanded-around' view that allows the capture of multimodal interactional processes in their phenomenal depth in time and space.