It may not require low-latency dynamic audio at all. It could be that you could transfer many ms of signal to the hardware in advance to ease your timing requirements.