It generate voices from a style and text. It’s an improvement over existing tts by adding tones and emotions.
Obviously it won’t replace a proper actor just yet, mostly because context matter, quality of sound is not to the standard yet, but sound upsampling is not impossible nowaday.
For a lot of non AAA project, with basic narrative of voice over, that’s already enough, especially if you go retro to match the sound quality, or use a context where it make sense.
It’s more a preview of stuff to come, the tech is here, and soon will probably be real time.
Think its pretty cool, especially as placeholders. That said, though better than many generated voices - its still pretty easy to spot. The second you do, it kinda ruins the character for me… its like, oh I’m just talking to a robot. I know reading tons of quest text etc. can be a bit much, but at least you get to do the voice work in your head… so to speak.
Interesting, and better than anything I’ve seen, but agree with others. It’s still very robotic.
I don’t consider voices to be one of those things where ‘oh that’s good enough, it’ll do’. Voices are not processed in the mind the same way that visuals are, it’s nowhere near linear in terms of quality vs enjoyment.
Enjoyable voices are a very precise thing, like sculpting a face where putting something a few millimeters to one side makes a world of difference. Even people with a good microphone voice are not good enough for voice acting, it requires strong emotional expression combined with very precise control of tone and inflection.
Personally, I’d either find some good voice actors or stick to text.
It will evolve anyway, I know it’s not there yet, I just find it promising given the pace in which such technique get perfected over time, the robotic aspect seems like it’s created at low sample, I wonder if there is other unrelated services to clean the voices and upsample them.
Anyway to not drown the forum in new thread from similar subject
Here a proof of concept for music
musenet improvisation on smb songs
and the latest in visual generation, this one isn’t artistically spectacular, to the untrained eyes it looks like a color filter (which they compare to later), but pay attention to the plant, down the line we won’t use shader no more, that’s certain
intel visual realism transfer networks
I signed up (you get 30 minutes free) and the price is good ($25 for 4 hours of generated voices). If you’re lucky you’ll get things off the first try but you might have to tweak things like annunciation and changing some things to use phonetic spelling to make it work.
It’s good but not great but better than doing it yourself. If you really need a polished end result there are other options (3rd party talent) but as Jason mentions in his video, now you have to manage that and find all the people that can do the voices. The end result is a little robotic at times but sounds pretty good and any voice can be good for an indie title as so few of them have them.
My only thought is that there are only about a dozen or so personas and voices so if everyone is jumping on here you’re going to keep hearing the same voice in different games. It’ll be like seeing Ethan all over the place again.
Still it’s a cheap alternative and a nice service. Definitely something you should add to your toolbox.
For sure. And it’s far better than it was even a year ago when I tried out their Beta.
I wonder who’s actually making it. Since I looked for one placeholder voice generator I’ve been bombarded for adverts for different versions of basically the same thing.
To me the main thing is that it might be useful for workflow. It means you can make sure everything in your game works - the right voices play at the right times, and the dialog makes sense to players, and all of that sort of stuff - before you find professionals at the end to make drop-in replacements so that your VO is actually, y’know… good.
Still available, free I think, though it needs a DAW to run in as its a plugin. Its not best suited to dialog, tbh, focussed more on ‘singing’ ie pitched phoneme sequencing a la Vocaloids.
