You are right about the delays. One of the real time systems could probably do it. Almost certainly the FPGA module could do the processing fast enough and it can be run independently from the desktop OS. I have not used any of the real time devices, so I am speaking theoretically, not from experience.
Since you are talking about a demo, perhaps you could set up a non-real time demo which shows the algorithms and processing, but works with recorded data.
Lynn