Wątek: Piper for NVDA

Strona: 2 z 2

21 z 28: Mckensie

There is a new version of this addon here which is more responsive.

25.04.2024 23:25

22 z 28: marchoffmann

Is this the beta everyone has by now? Or actually a new beta.
Miracles happen. The question though, is when...
26.04.2024 06:00

23 z 28: rudolf

looking at the version helps. 3.0.2, not 3.0.0.
#MCMacy4admin2024
26.04.2024 13:32

24 z 28: Musicant

Did someone try training voices for this thing? Share your experience: how much data used, which hardware the model have been trained on, how well is the result.

28.04.2024 18:03

25 z 28: marchoffmann

It's pretty intresting yeah. But didn't really have the possibility for that, I don't think it's quite easy, more for the nerdy side of the comunity.
Miracles happen. The question though, is when...
28.04.2024 18:28

26 z 28: Musicant

Well, after reading the instructions i came to the conclusion that Technically, it is more tricky than training a singing model like RVC but still quite possible as opposed to the tts technologies we had before. I definitely want to try it sooner or later but while i still didn't, i am curious to know other people's experience in that endeavour.
Even though the quality is not good and it is still more a thing to play rather than a ready-to-use tool for everyday job, the possibility of reading anything you want with almost any voice you wish in the nearest future is very interesting to me.

29.04.2024 04:28

27 z 28: x0

Tamas G has created a new google collab notebook for training this new real-time model. Or at least I think he created it, Musharraf might have done. He posted it in his mastodon @tamasg@mindly.social, so if you can bring up a search against his timeline for links you might find it,as well as some models he's been training like Keynote Gold, BNS Votrax and Mac Alex.
This mesage has been approved by x0.
03.05.2024 20:58

28 z 28: Musicant

Could someone post this link here for those like me who doesn't have this new fashioned social media tricky thing? Trying to train in colab would be a good idea while Google doesn't block this like they do with rvc.

04.05.2024 13:43