This is one of the best known features of the phase vocoder and works quite well compared to other methods. As far as my research goes the only implementations of the time stretch routine that are superior in sound quality is a neural net model.

To perform the time stretch the signal is encoded and a time scale vector is computed based on the scale factor. The signal is then interpolated against this time scale vector and decoded.

Original Signal

Time Stretched signal

M Files:
Example
timeStretch.m

Wave File Examples:
Wave File 1 - 2X stretch
Wave File 1 - 4X stretch
Wave File 1 - .5X stretch