With the rapid expanding of the integration of telecommunications networks, cable TV networks and the internet, the efficient and energy-saving video transcode service platform is in great demand. Such a platform together with workload prediction is proposed in this paper. A new accurate multi-scale prediction method is used to predict the execution time of transcode tasks in this platform. Special distribution and management scheme is applied to serve the purpose of energy optimization. With several experiments, it is proved that the proposed platform can be used to predict the workload of the transcode tasks accurately and save the energy efficiently by task scheduling optimization.