Generally, you will want to have your voices recorded before you start animating: that way, you can time the animation precisely to the lines. But it all revolves around what framerate you're shooting at: if you have a 2 second line and you animate at 15FPS, then you'll need 30 frames whereas if you animate at 12FPS you will only need 24 frames.

But how many frames does it take? Well, first you find out the total length of your line (you can do this in Audacity). Then, you multiply the amount of seconds it takes by the number of frames. For example, if you shoot at 15FPS and your line is 1.6 seconds long, then you will need (1.6 x 15 = ) 24 frames of animation to match the voice.

So now you know how much frames you need in total. But it doesn't stop there (at least, not if you want to have animation that actually matches the dialogue. If you have a line that says for example "Hey, that's MY monkey." then you will probably not want the minifig to have his arms flailing around like a madman. Instead, having him raise an arm at "hey" and then after that both at "MY" would be much nicer.

Well, you apply pretty much the same technique as before, only this time you find out how much time there is in between "hey" (movement 1) and "MY" (movement 2). After that you can easily find out how much frames you need in between these two points of emphasis/minifig movement points.

Then you have to mix the two together, but one usually does that with an editing program and not with the animation program, audio program (Audacity) nor a compression/frames-to-film processing program (VirtualDub).

And if you want a program to do it for you (calculate the frames), use WAVTracker. Its free. All you have to do is import the audio, type in your framerate and there you are.

Youtube