A rendering constraint is detected for a maximum time between capture of each a plurality of frames of a captured video stream of a consumer and a rendering of each corresponding frame of a consumer representation comprising a rendered three dimensional model rendered to depict a consumer appearance and a movement wearing a particular merchandise item. Processor elements accessible to the at least one computing system are selected that are able to generate the rendered three dimensional model within the time period specified in the rendering constraint. At least one processing task is directed to the selected processor elements to build the consumer representation comprising the three dimensional model and generate a rendering of the three dimensional model wearing the particular merchandise item as the captured video stream is received in real time at the at least one computer system.