I made an experiment some time ago. Not exactly a pure one, but still, I think it could be interesting/useful.
I intended to make a video comparison in the end, but ran out of gas. )
But I still made some images for comparison which would fit forum format, I think.
The end-result of the faceswap feels pretty decent... until you compare the quality with the source material. ) But that is understandable, considering the close-ups and the action.
This is how it started. 315k iterations Villain trainer
Next one is 840k iterations total
The third one is 1040k iterations total, but I also added more Tulsi Gabbard's faces
And the final one is 1300k iterations and I also added more Wonder Woman's faces
Now some of the comparison images:
I'm not 100% sure, but it feels like the nose moved a bit or got a bit shorter between the 1st and the 4th version. )