I find it’s worth clearly stating the accent you want (probably for each speaker separately). I usually end up defining everything. Leaving anything to chance more often than not produces random nonsense. One thing I’ve found almost impossible is having characters where one has one accent and the other one has another. The only way to do it seems to be separate clips using the final frame of one as the start of the next.
Voice intonation is hard to control, and not consistent. Also, when you’ve been trying to get a motion to work over 6-8 tries and when it finally works, the speech isn’t quite perfect, unless it’s unacceptable, I’m not going to keep trying.
I’m hoping to have the first NSFW Flicking Channels video ready later today. I should stress it won’t be full-on nudity (which is where I’m aiming to go) but the clips where there is unintentional nudity (where the AI made a mistake), such as missing clothing, nip-slips etc. There will be occasional full-frontal nudity in them but not a lot.