Source devices, softwares and methods are provided for predicting at a transmitting source how well a specific encoded frame, relative to other frames in a stream, would be reconstructed if it were not available for eventual playout at a receiving destination. If poorly, then the specific encoded frame is packetized and transmitted advantageously with respect to other packets in a stream. The transmitted stream will have higher jitter. Predicting is performed at the source by synthesizing a replicated frame, which emulates how the specific frame would be reconstructed at the destination. Then the replicated frame is compared with the decoded specific frame. The eventual playout will have a higher fidelity to the original, since reconstructed from more critical frames. Receiving devices, softwares and methods are provided for receiving in a jitter buffer a stream of packets containing encoded data frames, determining that they have been packetized differently from each other, and setting a playout delay of the jitter buffer responsive to the determination, regardless of high jitter.