The run-time power consumption model for multimedia application routines in an embedded system is developed in this work. A wide range of benchmarks for these routines such as image, video and speech coding algorithms are examined extensively to characterize their power consumption behavior. Then, based on this study, a run-time power consumption model is proposed to achieve fast and accurate power usage estimation, where the model depends on the specific application routine involved as well as the applied voltage and the operating frequency. It is demonstrated that the developed model is able to predict the cumulative energy consumption of validating sequences with little error. Finally, this model is validated with a commercial VLIWprocessor by checking its actual power consumption behavior.