What makes you reject an AI lip-sync result?

I'm on the FreeLipSync team. For creators using lip-sync on their own videos, what is the first thing that makes a result unusable: mouth timing, changes to the face, or unnatural movement?

If you have a difficult example, what happens in the clip—fast speech, a side profile, a beard, or a hand covering the mouth? A description is enough; no need to share private footage.

Which issue would you want fixed first?