I have never tried this, but I suspect it is possible to create a "motion envelope" of the new gear geometry. The "problem" in this case, is that the motion envelope is fixed relative to ground. So you will have to define the motion of your parts so that one of them, "female" is fixed to ground, and the other one has its motion defined relative to the fixed part. (*testing*). It turns out the resulting geometry (I would think), is too coarse, see below. Note that the resulting "toothed" surface is the negative of the desired gear wheel.
Motion model used:
Next attempt would be to create a "trace curve" from the "male" onto the "female" paper part. In this case you can prescribe motions to the parts as you wish. This only creates the profile of a single gear tooth profile. I guess you might need to create several trace curves to get an accurate femate tooth geometry. Material on the "inside"of the trace curve(s) should be removed. It might be necessary to generate several trace curves, for a series of datum points along the "male's" gear surface.
I've done this not using Mechanism, but by creating a surface of the 'male' tooth and then patterning it at very small increments. The increment size was limited by accuracy / regen issues so the result was a bit facetted, but we actually milled parts (face gears) from the model and they inspected OK and seemed to work fine (in a lightly-loaded actuation application though, not 'proper' power transmission).
