In this talk, we present the design, implementation and evaluation of a utility generator that plays an important role in a utility-based adaptation framework by dynamically generating bandwidth utility curves through the passive process of inspecting video content on-the-fly. The utility generator interacts with a network adaptation mechanism to periodically re-negotiate bandwidth to match the generated utility curves and comprises four elements: a profile selector, distortion meter, utility assembler and utility stabilizer. Using a set of MPEG video traces we analyze the performance of the utility generator and discuss the process of utility stabilization, which represents a technique that keeps the generated utility curves meaningful over longer periods approaching the network adaptation time-scales. We demonstrate through experimentation that utility generator represents a promising new approach to delivering scalable media over content-based networks.