In this paper, we propose a key-frame- based error resilient coding scheme for video transmission over differentiated services networks. Key-frames are fixed in advance. Each frame, if inter-coding, can only choose the latest coded and reconstructed key frame as its reference picture. After coding and packetisation, compressed video packets are transmitted with differentiated service classes. More specifically, we assign the key-frame packets to the assured class, and assign all the other packets to the regular best-effort class. Although the assured class has much lower packet loss rate than the best-effort class, it cannot guarantee no packet loss. Retransmission is used on the source-layer level in the case of losing a key-frame packet. By this means, we can stop transmission errors propagating along the prediction loop. Our scheme can not only be used with on-line coding video transmission, but also can be used with precompressed video transmission. Simulation results verify the validity of our scheme.