AI voices still sound drunk

I got the Maki AI English Pro, and have been working with the auto-pitch tuning. It just sounds drunk and off-key. I have pitch-correction on and improvisation at 0, and she still sounds like an off-key mess. I see the custom style button, but it’s always greyed out and I have no idea how to use it, and if it would help.

Does anyone have any tips on how to make her auto-tuning sound better? Thanks!

Something you can do that’ll help is to set the Improvisation down to 0. That’ll still let the Auto Pitch Tuning work without any extra, well, unnecessary Improvising. From there you can add extra [ - ] notes to mold and reshape how the software handles the pitch changes between the extra notes.

Randomizing the Seed will also change the results a little bit. Keep hitting Randomize until there’s some Auto Pitch Tuning that sounds better!

Hope that helps!


Neat, I’ve been doing all of that, except the extra notes. I’ve never tuned vocalsynths before, so it’s exciting! (read: terrifying)


I personally think that having pitch correction on makes it worse. It sounds weird to me, like it’s trying to force someone to sing on key and it disrupts the transitions between lyrics.

For Maki, I like to set the Improvisation around 35-40, and then I also highlight all the lyrics and set the vibrato depth to around 40.

She will still have odd things pop up here and there, so I always do a listen run through so I can manually fix any bizarre pitches that crop up.

I like that! Thanks!