In this chapter, we present a function-free method for the offline learning of the perceptual quality of real-time interactive multimedia applications (RIMAs) that are defined using a single metric and simplex control. We introduce a Just Noticeable Difference (JND) surface to relate perceptual quality in awareness to scenarios evaluated by subjects under specific network and operating conditions. We hypothesize the monotonicity of awareness and present a dominance relation that enables pruning most subjective tests, as their resulting awareness value is likely within a prescribed error tolerance of the value from tests already conducted. By strategically selecting test scenarios and interpolating the results of previous subjective tests, we can determine the perceptual quality of all the operating points on a JND surface. Our function-free method represents a breakthrough compared to traditional psychophysical function-based methods, as it enables us to determine the entire JND surface within an error tolerance by conducting only a small number of rounds of offline subjective tests. Using this solution as a foundation, the next chapter derives the composite perceptual quality of multi-metric, multi-control applications.