Winkler et al. (2003) demonstrated that it is possible to evoke MMN with natural stimulation (well, almost natural). The sound and the illustrations below are courtesy of Istvan Winkler.
The subjects were busy watching and listening to a movie. They also heard background noise from the street, over which 11 steps were heard. Either the second or the one but last step were different from the others, and the MMN was calculated as the difference between these two conditions - presumably, when the different step was at the end of the sequence, it was detected as a deviant relative to the others.
In the illustration below, the footsteps are plotted in blue, the street noise and the movie in gray. The deviant step is at about time 8 s.
Here's the resulting MMN:
The sound example below is the deviant condition: