Any Auto Lip Sync tips for more accurate viseme selection?

I'm trying to use Auto Lip Sync on a bunch of character animations.

I've been supplied audio for each, some male voices, some female and some 'character' voices (think higher pitched female/child voice). They're all good clear recordings.

The auto lip sync works, but in many cases, usually the higher picthed/charcater voices, Animate choses a random bunch of visemes, making the lip sync look totally random.

Male voices seem to work better. I thought therefore that it migt be a pitch thing, so tried editing some of the female/character voices to be lower pitched and tried some tests, but while the results were better, they still were not great.

Anyone found a magic formula to get better (if not perfect) lip syncing?

Might it have something to do with the audio format? Or is it just that some voices are easier for the AI to 'hear' than others?

Currently, some of the auto lips syncs are so bad that I'm basically having to go through and do them by hand anyway.