Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

WIP 🎄 Text-to-speech in VAM (Patreon)

Published:
2022-12-24 00:09:55
Imported:
2026-06
haven't archived this post yet. have a subscription? use the importer!

Videos

Content

VaM can feel quite empty & silent at times so a feature of the new UI I'm working on is a personal assistant of sorts that can talk. It uses the locally available voices, not 3rd party online services.

Don't get too excited though, some of you will not like it! I'm going more towards a sci-fi ship AI kind of vibe rather than giving voice to models and going for the realism feel many seem to want from VAM. Lipsync text-to-speech is possible, it's actually the first vam thing I shared. However it required installing & running a bunch of complicated software, most people wouldn't even bother, and in the end it looked kind of meh and slowed down the responses considerably so I'll avoid that direction, for now at least

Files

Previews only

WIP tts vam

This is "WIP tts vam" by xspqrx on Vimeo, the home for high quality videos and the people who love them.

Comments

Tuckwit

Yea, come on Microsoft. Instead of Hazel or Susan can we have a 'Sexy Saucepot' voice as well please! Seriously, great news though. It's still better than some other options. Apparently Azure is free but I suspect there is a limit on how many words can be done a day, much like google and any others. Also I don't really want M$ or Google analyzing/recording all the naughty sentances we send them! And of course we already have AdamAnts Realtime Lipsync which does a decent job to the lip stuff anyway

SPQRAeternum
>>98149852

big corps are doing the opposite in this field unfortunately. there used to be small companies making high quality offline voices, you'd pay once and have them forever. They got bought off and shut down. Ivona was the leading one with fantastic quality voices, Amazon bought them 10 years ago, closed everything and turned it into their cloud service amazon polly. To this day if you look up best local tts voice people will say it's Ivona Amy or some other Ivona voice that was made over 10 years ago. If anything comes remotely close, a big corp writes a check and adds it to their cloud pay-per-word service. Ivona voices can still be purchased though from 3rd party resellers (deskshare, nextup,harposoftware) that still have the original ones before amazon's kiss of death. They're like $30 a piece and totally worth it. There's also AI voices coming up but they're a bit tricky to use, they're not integrated in the OS as system voices

SPQRAeternum
>>98149852

this won't work with Realtime Lipsync, that plugin was what I was referring to with lipsync to work with that you need an audio file for each speech. that's doable but it involves a lot of geeky stuff, running a local server and using a software that saves speech to files. I don't think more than 5 people would even bother installing all that

thomas d

I think there was someone else working on something similar I saw recently, I think you had to download a program to do it, then convert to wav or what ever. Some creators just write up what they want and get a voice actress to speech it to a wav. I even saw a post about text to speech using celebrity voices. Anyway Audioharlot is my favorite voice actress. Vihper uses her and maybe other creators. She has quite the talent.

Anonymous

Would be cool to link the "replika" AI chat bot app into Vam with lipsync. If you've never seen it go to your app store on your phone and install replika and check it out. Would it be easier to link something that already exists it rather than build from the ground up?

SPQRAeternum
>>98303367

https://replika.com/legal/terms point 6.3 besides not being allowed to, it's a bit of a dead end imo - not practical. while the AI is amazing, you can't really link it to vam, there's no api. If you do a split screen and put replika now on one side and vam on the other, it would be the exact same thing except that it would be a delay, you'll have to install and run extra software and the model's lips will move to the sounds. At most the model might smile if you detect the word "happy" and small things like that to sell it, to me that's boring and not worth the trouble but there's somebody else already doing something like that if you're interested https://www.patreon.com/posts/75433944