This paper investigates deploying a village level community-radio application on top of a MANET comprising completely of basic mobile phones. We envision a system, where any user in the network is equally empowered to generate and distribute audio content to the entire network, using his or her mobile phone. The paper focuses on the study of suitable broadcast algorithms for the network. In this context, we propose a novel broadcast scheme where nodes bank on their memory to decide whether to forward packets of an audio stream-capitalizing on their past behavior to stabilize on fixed routes for the entire stream. In our scheme called Environs Aware Broadcast Mechanism (EABA), a node gauges the local mobility around itself, and uses that to decide which broadcast mechanism to use. When mobility is high, it uses SBA (Scalable Broadcast Algorithm), a popular neighbor -knowledge broadcast algorithm with high overheads, but when mobility is low, it switches to MaBA (Memory-aided Broadcast Algorithm). Extensive simulations on a village-level MANET, confirm that EABA is successful in substantially reducing jitter, latency and packet loss: all critical metrics for an audio application. At the same time, EABA does not incur other overheads and maintains the same levels of reachability and efficiency as SBA