For what i've seen, fast npc interactions without a reply option end up with audio starting right away and video with subtitles doesn't catch up and they load slow, ending up sometimes with just black video.
The best example is the woman with broken cart at start of act 1.
Theorically they should be synchronized and videos be able to load right away to not let us wait and take coffee every time we interact.
I've a pretty powerful computer, my only possible issue is that i use a HDD and not SSD to store games, but that shouldn't be an excuse.