Making a new voice for it to sound 100% like the old one is, by definition, completely pointless. It's no surprise you hit a wall in the training data part if you're trying that, because only V1 Neuro sounds exactly like V1 Neuro.
The whole situation looks like an open alpha game that has no updates for 10 years because "the project we were developing was not to our standards, so we're remaking the entire game for the 4th time". Please don't fall for that trap, it's been 2 and a half years already (since V2 voice, that was supposed to go to neuro)
I know it's your project at the end of the day, but since this is a project destined to the public, try not to 100% isolate yourself on your own bias, specially if you're trying to seek the impossible. You already got an overwhelming approval from the several selected people you handpicked, as you said on a dev stream. Please at least listen to their feedback.
I would not be opposed if you released the V3 voice and it only sounds ~90% like V1 Neuro. The worst thing that could happen is the public saying that it's shit (which is what you're doing to yourself already, so it'll make no difference and you'll be validated).
Please don't let your seek for perfection harm Neuro's development



I'd rather wait than have her sound exactly like evil. (the bloopers sounded like from an abber demon stream, though we don't know how long ago they were from)