A serverless peer-to-peer (P2P) multi-party real-time audio communication system and method in which each of the peers takes a turn mixing and compressing the audio and redelivering the compressed audio. An input audio stream is divided or split into frames. At each frame, one peer node is selected to mix and redeliver the audio to the remainder of the peers in the network. The number of frames mixed and redelivered by a certain peer is proportional to its available resources (such as the upload bandwidth or computational power). The P2P audio communication system and method flexibly balances the load of the peers, such that peers having more resources are able to assist peers having fewer resources. This enables the P2P audio communication system and method to conduct a multi-party audio communication session without the need for powerful servers or peers.