[Marc Newman]

I was following along and he brushed over a step. But, I’m not going to repeat how it was explained. Instead, I’ll explain what you’ll have as a final result:

One precompiled composition with three layers (in order, top to bottom):

1) A solid color. The actual color doesn’t matter. The point of this layer is to contain one or more masks. The way in which is follows the movement is by setting the parent of this layer to the layer with the transform moves. Note: You could apply the tracking (transform) info to this layer, instead of using a parent, but you don’t really want to do that because you won’t be able to use transform for anything other than this layer.

2) The parent null object “Tracking” layer. This layer doesn’t even need to be visible. It’s just here to provide instruction to the top layer. After applying the tracked motion to this layer the first layer (including the mask) will move, trench, rotate.

3) Your image that will be within the mask (#1). You don’t have to apply any effects to this layer, as it will probably be better/easier to apply the effects in your original composition.

You will also have your original composition, prior to making the precompiled comp. Your original composition will contain (top to bottom):

1) The precompiled composition.

2) The background / outer masks/layer(s). Basically, everything else.

Apply your adjustments to #1, above.