some of the voices don't show up on the list, like Vex from skyrim or Elizabeth from bioshock! plz fix!

Originally posted by Korbe:

some of the voices don't show up on the list, like Vex from skyrim or Elizabeth from bioshock! plz fix!

Vex is in the list of "Get more voices".

Elizabeth is on Patreon, Bioshock does not have a category mod on Nexus Mods just like many others.

---

OP post is garbage. Comment

Originally posted by Pendrokar-OD:

Originally posted by Korbe:

some of the voices don't show up on the list, like Vex from skyrim or Elizabeth from bioshock! plz fix!

Vex is in the list of "Get more voices".

Elizabeth is on Patreon, Bioshock does not have a category mod on Nexus Mods just like many others.

---

OP post is garbage. Comment

Editing old lines once saved resets every bar.
The spacing between every bar has inadvertently touch radius almost as if this was designed for an iPad.
The voices I played with had a hard time pronouncing S's and I had to misspell words for it to sound normal, of which, resets every other word you worked on even if it's just to edit one word in particular.
There's a slew of others, but the application is bugged, simply, I felt like I was fighting with it for three hours only to get one line. I had to keep restarting because functions would break, especially when I increased the speed or lowered, it literally would just ♥♥♥♥ out an error code and then would be cursed in terms of pace afterward. Other than that, you're a fetal alcohol syndrome baby.

Now that's better.

the thing is, this "ai" nonsense is actually just butchering small sample sizes for bottom tier vocoders. TTS and vocoders are like 70 year old tech, but the thing about a GOOD vocoder (thats the file that handles the actual voice, not the program) is that it literally requires hundreds of hours of audio. to get GOOD quality voice clips from TTS, you need at the bare minimum a few dozen instances of every syllable, ideally you need at least 300. ive worked with (not messed around with, but done actual get paid work) vocoders on and off for about 20 years, and it is a slow laborious process. what the "ai" programs are doing is what lazy vocoder techs do : take a single sample or small number of samples, chop up the sample into individual sound clips and start stretching and filtering the individual clips to obtain different sounding pronunciations of the syllables. the end result is something that sounds *almost* right, where as doing it the correct way results in a TTS that is indistinguishable from a human 99% of the time.