Gotcha.
You need to import or draw the mouth shapes -- one for each layer and one frame in length, sequentially. In other words, add all the "images" to a group. Go layer by layer, starting at the start of the project and the bottom of the layers list: type I, O '->' (right arrow) (up arrow) [make sure the next layer up is selected]. Continue with each layer until you have each layer at one frame length starting end to end.
Select the Group all your images are in and type K to clone it. Turn off the group (if necessary).
Replicate the Clone.
I set up the animation with a Line and set all the start and end points to 0 (a zero length line). (I accidentally left the "points" at 5...
not that it matters here 😉 )
In the Replicator controls, deselect Play Frames.
Keyframe the Source Start Frame to match the audio.
Since you'll be jumping over various shapes back and forth, right click on the Source Frame Offset and Show in Keyframe Editor. Select All the points, then right click on one of them and set the interpolation to Constant.
I put together a very short demo here
http://youtu.be/eJcIsM1Ucqw
and you can download the demo project so you can take it apart here:
http://sight-creations.com/fxexchange/CartoonSpeechDemo.zip
I used a copyrighted image... so don't pass it around. You'll notice I basically just masked and moved the single image to create the basic seven frames of the animation. I didn't attempt to line everything up (something you should do though!)
Can you guess what the cartoon is saying??