Pipecat is an open-source framework that orchestrates voice and video streams while linking speech-to-text, large language models, and text-to-speech components to minimize latency in AI applications.