Integrating voice communication into a game console to minimize or eliminate voice data processing by a primary processor. Input voice data from a microphone or a network is processed by a secondary processor and stored in a circular buffer. Drift between storing and reading the processed voice data may result from differing data rates, interrupts, and other latencies. If the circular buffer, accumulates an amount of data that exceeds a predefined threshold corresponding to a human perceptible latency, a pointer in the circular buffer is reset, so that only a portion of the processed voice data is output. A stream of packet contexts each indicate a location and length of voice data in the circular buffer to be output. Preferably, the output voice data is encoded in a standard digital format, such as universal serial bus. The output voice data may be communicated to a network or a sound transducer.