After performing several face swaps and even looking at some examples from the thread "choosing faces that match" it seems that the current models all have a pretty big problem.
The models I have seen do a good job swapping the eyes, nose, lips, and cheek bones but the chin, jawline, and distance from ear to ear do not change much if at all. This makes it pretty hard to perform a convincing swap without finding a nearly perfect match for the above listed areas.
I am not the first person to notice this 'problem' and I have read that models are being developed to address this issue.
I have read that lae based models do a better job at replicating face shape at the expense of face detail.
Other forks of deep fakes are using 'morphing' models to address this issue.
This reminds me of GPS technology in the 90's. It's pretty good nowhere near its potential.
What is the best current model faceswap.dev has to replicate the whole face? I know there is an option for head replication but this really only works in certain circumstances (mainly for people without hair).
Is this an issue that we are just going to have to wait for the technology to be allowed to evolve?