The present study tests whether presenting video modeling examples from the learner's (first-person) perspective promotes learning of an assembly task, compared to presenting video examples from a third-person perspective. Across 2 experiments conducted in different labs, university students viewed a video showing how to assemble an 8-component circuit on a circuit board. Students who viewed the assembly video recorded from a first-person perspective performed significantly better than those who viewed the video from a third-person perspective on accuracy in assembling the circuit in both experiments and on time to assemble the circuit in Experiment 1, but not in Experiment 2. Concerning boundary conditions, the perspective effect was stronger for more complex tasks (Experiment 1), but was not moderated by imitating the actions during learning (Experiment 1) or explaining how to build the circuit during the test (Experiment 2). This work suggests a perspective principle for instructional video in which students learn better when video reflects a first-person perspective. An explanation based on embodied theories of learning and instruction is provided.