With the rapid development of new media video services today, traditional standalone video transcoding capability has been a bottleneck. This paper proposes a new video transcoding scheme based on the research on Hadoop cloud computing platform and the current mainstream audio and video processing tool FFmpeg. This scheme fulfills distributed transcoding method by using two core components in Hadoop: HDFS(Hadoop Distributed File System) and the programming ideas of MapReduce. Meanwhile, the paper also describes in detail the specific processes and design of distributed transcoding. Finally, experimental results show that The distributed transcoding scheme has greatly improved the efficiency. Simultaneously in the experiment, the segment size of each video file also impacts the time of video transcoding process. The experiment shows that as the segment size goes from small to large, time consumed by the transcoding process index experiences a curve: firstly it changes from large to small, then drops to its lowest point and gradually rises. The lowest point exists when the segment size is adjusted to 32M.