#vasynth

TL:DR COMPLETE REVOICE OF THE PLAYER CHARACTER USING COMPUTER MAGIC

Now this... THIS is the project I've been working on that's kept me from starting my new game properly. I wanted to have this in from the start and it's finally gotten to... I suppose a usable state?

So i wanted to post a bit to show it off, and to ramble out some thoughts on it.

Some of you have maybe heard a bit about AI powered voice synthesis, and there's a developer (DanRuta) who's made an actual accessible implementation of this tech for Average Computer User.

I've been following the project since fairly early on and there's been huge leaps and bounds in the tech. It's really cool stuff. Far from being able to completely replace a human actor (not that I would ever want to) but it's a great supplement.

I ALSO saw the opportunity for something to exist that I only ever vaguely fantasized about, considering the scale of this game (15000+ player voice lines, depending on how much of the DLC you install.)

So with a custom voice model trained, I started on this a few months ago.

All of Kate's voice lines here are completely original, generated from the xVASynth software. It works similar to Text-to-Speech, though more sophisticated. I extracted all the player lines from the game (and all my mods...), processed and cleaned those, and fed them through the software's batch mode. (Which takes all night to generate for the base game, whoof)

More hours then spent sifting through a large sampling of lines to find inaccurate pronunciations and correcting those - and then regenerating the voice lines overnight again... Even though the software is doing the audio work I still had to put in a ton of time and effort. Less, though, then would be needed to record every single voice line in the game.

The results are still a little imperfect and a little artificial. The trade-off for doing things this way. But it's made something possible that I couldn't have done otherwise.