No we use ML models that are based on single deconstructed images.

No we use ML models that are based on single deconstructed images.

The lipstick model is as simple as identifying facial mouth landmarks and applying a color filter within those bounds.

The glasses model deconstructs a single image into 5 parts and scales and moves them independently so that the head can be angled and the images can be superimposed independently.

This creates a more dynamic model. Albeit it requires a bit more tuning to be more realistic.