A system and a method for producing a low-latency video for transmission over a network. The low-latency video may be created by modeling a select portion of original video data and comparing a current frame of the model against previous frames of the model in order to estimate the select portions of the original video data. The estimated, select portions of the original video data may be combined with a remainder of the original video data (such as background images) in order to produce the low-latency video. Model data and the original video data can be transmitted over the network using different paths in order to ensure that the model data is transmitted as quickly as possible, thereby allowing enough time for a morpher to process the model data before combining the model with the original video data.