A wireless Mobile Ad Hoc Network (MANET) comprises of a number of mobile nodes that uses multi-hop routing to provide network connectivity. MANETs require self-organizing capabilities as there are no centralized points (base stations, access points etc), and each mobile node functions as router and/or hosts. The wireless topology in MANET can change rapidly with mobility of nodes in unpredictable ways or remain static for long periods of time. MANETs have applications in neighborhood area networks (NANs), impromptu communication among groups of people, disaster management and dynamic military systems. As progress in MANET continues, there is an increase in demand with regard to supporting content-rich video streaming in such networks. This is due to the fact that real-time video is far more substantive than simple data communication. This work involves implementing a Genetic Algorithm (GA) based multipath routing methodologies in a proactive routing protocol (Optimized Link State Routing Protocol) to send/forward/receive multimedia streams on experimental testbed. We study the problem of multipath video routing in wireless ad hoc networks by following an application-centric cross-layer approach. A full implementation of GA-based routing and real-time video conferencing application (server and client) written in C++ is presented. The robustness of our routing scheme was tested through experiments using five computer nodes. The performance of the routing protocol for video, as well as issues such as applicability and scalability in practice are addressed.