> b) anybody has some idea how to do that?

*Queue* up the data.  Wake up the upper level when you reach
the low-water-mark while sending data (i.e. playback) and
high-water-mark while receiving data (record).  Low-water-
mark == nearly empty buffer, high-water-mark == nearly full
buffer.

Since this is sound, you can predict when these events occur
so use timeout() to make these calls at appropriate times.
The buffer should hold enough data to last a couple seconds.

Ideally this should be done using Streams.