Abstract: A method and apparatus for performing the asynchronous reconstruction of a digital data stream is disclosed. An input data stream having an input rate, such as digitized speech, is stored in a circular buffer. Within the buffer, read and write marker positions are determined. Upon subsequent retrieval of the data stream at an output rate, data is taken from the position of the read marker each time data is written to the position of the write marker. Of importance, the read and write marker positions are determined such that they provide a delay greater than the difference between the input rate and the output rate over the length of the buffer. The data stream processing is then modified by ignoring at least some of the data stream when the input rate is faster than the output rate, and repeating at least some of the data stream when the input rate is slower than the output rate. In this manner a method of digitized voice reconstruction without reliance upon network synchronizatin is achieved.