Audio restoration Tips

[Sunnyl75]

I'm trying to use a trained model of a speaker from well recorded sources to restore damaged or low fidelity recordings of the same person.

The timbre of the voice and the quality of the audio transfers extremely well, however I've found that the output audio tends to stray from the input audio.

So far I've noticed that:

• the emphasis of words in a sentence sometimes changes
• t's, b's p's and d's in particular are quite weak, and s's are a little lispy
• The pitch of some words jumps up in a surprising manner

Are there any settings I could adjust to make the output audio more closely align to the input audio apart from the timbre?