This version was epoch 20 of a dataset with 100 videos. Seems to work best with a cfg of 7 or 8.
This model aids in generating clips looking down a woman's blouse.