Aerial cinematography is significantly expanding the capabilities of\nfilm-makers. Recent progress in autonomous unmanned aerial vehicles (UAVs) has\nfurther increased the potential impact of aerial cameras, with systems that can\nsafely track actors in unstructured cluttered environments. Professional\nproductions, however, require the use of multiple cameras simultaneously to\nrecord different viewpoints of the same scene, which are edited into the final\nfootage either in real time or in post-production. Such extreme motion\ncoordination is particularly hard for unscripted action scenes, which are a\ncommon use case of aerial cameras. In this work we develop a real-time\nmulti-UAV coordination system that is capable of recording dynamic targets\nwhile maximizing shot diversity and avoiding collisions and mutual visibility\nbetween cameras. We validate our approach in multiple cluttered environments of\na photo-realistic simulator, and deploy the system using two UAVs in real-world\nexperiments. We show that our coordination scheme has low computational cost\nand takes only 1.17 ms on average to plan for a team of 3 UAVs over a 10 s time\nhorizon. Supplementary video: https://youtu.be/m2R3anv2ADE\n