A server (101) stream-transmits data encoded by inter-frame prediction. The server (101) includes a transmission data generation unit (116) and an output unit (114). The transmission data generation unit (116) adds to a key frame which can be solely decoded, at least one difference frame following the key frame, so as to form one transmission unit data and generates buffer key frame having information on one transmission unit data including a frame number. The output unit (114) transmits the buffer key frame generated by using a key frame to be transmitted at the transmission timing of a key frame and transmits an arbitrary frame at the transmission timing of a difference frame. This can reduce the buffering time when switching the terminal for reproducing the stream between a plurality ofterminals and smoothly perform the switching.