Modding, Mission Design, and Coding > FS2 Open Tools
Suggestions for TTS Voice Acting Software
Trivial Psychic:
I am strongly considering using pre-recorded TTS files for my overhaul of Den5's TBP campaign "In The Beginning", aside from voice files captured from the movie itself... of course. I learned from Admiral Nelson that for the recent SAH overhaul, Replica Studio was used, and while I am intrigued, I was also hoping for a more cost-free alternative. Can anyone point me to some cost-free alternatives?
I was recently investigating Balabolka as an option, but after installing the program, I am puzzled to find that it only recognizes 3 of the 13 Microsoft English voices that I've downloaded. I've sent an email to the only contact link I could find on the page, with hopes that it will lead to someone both willing and able to provide useful feedback, but if anyone here has experience with this program and could help me out, it would be appreciated as well.
themaddin:
Festival (http://festvox.org/) is a free speech synthesis system with several voices working best on Linux.
A demo including several voices is online here: https://www.cstr.ed.ac.uk/projects/festival/onlinedemo.html
A good starting point IMO is the archwiki page https://wiki.archlinux.org/title/Festival
or the (German) Ubuntu wiki page https://wiki.ubuntuusers.de/Festival/
Some time ago, I wrote a script and necessary engine functionality (thread)
to integrate festival into FSO at runtime. I abandoned that project as unsuitable, mainly because speech synthesis
on longer messages within message chains took far too long to be useful.
Also, manually adjusting speed and pitch of the generated speech will certainly allow much better quality,
and 'terms of art' are painful to listen to unless the correct intonation is annotated.
As much as my time allows, I am interested in contributing to a workflow for creating such voice files,
including the setup of a WSL image with the necessary tools, provided enough time and suitable upload space.
Trivial Psychic:
The Balabolka program creator responded personally to my inquiry, and provided me with the proper batch file to activate all of the voices. At the very least, I can use this to create some placeholder audio files, and then replace them with something better quality down the line if I want.
Darius:
I've made heavy use of CereProc
https://www.cereproc.com/
For UK accents which is quite handy, most stuff out there tends to be US accents. Just use the embedded demo tool in the browser.
https://ttsmp3.com/ gets used a lot too for Spanish/European accents. Just need to disable ad block on their site to use but it's worth it i find.
I've recently started using ElevenLabs voice Cloner/voice synthesiser program, which is browser hosted with processing done on their side. It's quite high quality and if you haven't got many lines, quite doable without paying anything.
https://beta.elevenlabs.io/
If you don't mind putting in your credit card details you'll get the usual "1 month free, cancel after 1 month" starter subscription but it's high quality and a decent enough price that I may continue my subscription just to have a go at creating custom voices.
I've used it for several voices here.
Trivial Psychic:
Thanks for those suggestions, Darius. I've been continuing to use Balabolka so far, despite a few mildly irritating glitches, and limits on the number of voices available to me.
I tried those 3 sites you provided. CereProc doesn't seem to have a download option in the test box. TTSMp3 does have one, but both this and CereProc lack some fine-tuning controls I've used in Balabolka, primarily the ability to chose reading speed and pitch. You can also adjust these down to a per-word basis, which I've been using to emphasize words to make my lines more than a simple read-out. On the other hand, ElevenLabs seems to figure that out on it's own to some degree, thanks to the fact that the test line I was using, employed BOLDING on words I wanted emphasized. Eleven got it and used it, the others didn't. The only disappointment was the lack of test voices at Eleven Labs, though I'm guessing that there are a LOT more available if I went for the Demo itself.
I think I'll stick with my original plan of using Balabolka for now, and then maybe graduate to something a little more professional once I'm more certain about the dialogue. ElevenLabs DOES seem like a front-runner for that though.
Navigation
[0] Message Index
[#] Next page
Go to full version