While a number of commercial imaging systems are capable of acquiring images with very high frame rates, they are usually designed for burst operation, storing data on internal memory which is later read out at speeds much slower than the experiment duration. This results in essentially blind acquisition, and, moreover, the total acquisition time is limited by the available memory, which is usually not enough to cover the full duration of a dynamic process. Our developments solve this problem by streaming the data from the camera directly to a dedicated server, which is able to make the data available for previewing and analysis essentially in real-time, while the amount of data that can be aquired is limited only by the server memory or the attached file storage capacity. This opens up countless new opportunities for the observation of dynamic systems with high temporal resolution, such as crack propagation in materials, bubble growth in metal foams, musculoskeletal motions in insects, and many more. Making use of a modular and easily extendable, parallelizable architecture, the same data streaming approach can be used and adapted for other high performance detector systems.