We propose a method for the time scaling of digitally sampled audio signals using a three part signal model consisting of sines+transients+noise. The three part model provides an accurate and flexible parametric representation for a wide range of audio signals. Because the proposed time scaling method manipulates each of the model components separately, the method allows modified tonal components of the signal to preserve pitch, transient components of the signal to preserve edges and noise-like components of the signal to remain noisy. The method therefore provides robust and natural sounding time scale modifications for a large variety of signals.