Abstract Video transmission over unreliable networks presents new challenges for research in video coding. In fact, in video coders that use predictive coding schemes, even a single packet loss may cause errors during the decoding process propagating in time. A way to increase the robustness of the coded stream and reduce the length of error propagation is the use of a multiple description coding (MDC) technique. The aim of this paper is to apply a novel sequence-based error-concealment (EC) algorithm to an unbalanced MD video coding system that generates both a high-quality (HQ) and a low-quality (LQ) description. In order to recover a loss in the current frames, the EC algorithm takes into account not only the spatial/temporal neighboring of the region the lost data correspond to, but also what is going to happen in a significant number of future frames looking both the HQ and the LQ descriptions. The sequence-based EC algorithm presents good performances when applied to video streams. The comparison with concurrent error-resilience techniques shows interesting results.