As a gearbox/transmission designer, I agree with Elvind that for most purposes it's not necessary to model the teeth accurately, as the form is generated by the cutting machines, and they take the gear data in parametric form in any case. Most of our gear models just have an arc to represent the flanks - it looks good enough, is easy to model and regenerates quickly.
Gear calculation is best done by purpose-written software (such as Ricardo SABR-GEAR: http://www.ricardo.com/en-GB/What-we-do/Software/Products/SABR/Gear-design/</shameless plug> Image may be NSFW.
Clik here to view.
).
Having said that, there have been a few occasions where I've wanted an accurate model to run FEA on. I've found that the simplest and most flexible method for spur gears is to use an equation-driven curve to generate a single involute, which requires just the base circle diameter. You can then take the over-pins dimension to generate a sketch making the pin tangent to the flank, mirror the involute using a plane through the centre of the pin, and then use another sketch to create the root fillet tangent to the root diameter.
Having said that, this assumes that a full-fillet root is a sufficient representation (strictly speaking the root should be a trochoid); and I've yet to create a helical gear successfully - controlling the normal direction and rotation of the section correctly as it sweeps along the helix is difficult to say the least.
