User avatar
Tamas G @Tamasg@mindly.social
5mo
Still sad, but I'm still workin on NV Speech. Did not abandon it. I've fixed all clicks in the driver, and the next version? It'll have a sample rate combo-box. I am thinking some people will like it at 11025 HZ but some might want 22050. This one will let you pick. It's important for sound quality and the default 16000 may sound too harsh on some ears who are used to Eloquence's 11025.
2
2
1
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
5mo
@Tamasg Hah. Sample rate changes were the next thing I was going to suggest. We think alike, apparently.
1
0
1
0
User avatar
Tamas G @Tamasg@mindly.social
5mo
@fastfinge yeah, try it at 11025. I think it removes a lot of the sharp edges while still not sounding too bad. 22050 to me also sounds better than 16000, 44100 is just awful. But people will now have the freedom to choose.
1
0
1
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
5mo
@Tamasg It's amazing how different 11025 sounds. Also makes Adam sound way better, and benjamin way worse.
2
0
0
0
User avatar
Joshua @J3317@allovertheplace.ca
5mo
@fastfinge @Tamasg how do you change the sample rate? i updated the addond but can't find it in nvda settings or the speech settings ring
1
0
0
0
User avatar
Tamas G @Tamasg@mindly.social
5mo
@J3317 @fastfinge oh it's not in the ring, just a combo-box next to volume kinda like how IBMTTS had it.
1
0
0
0
User avatar
Joshua @J3317@allovertheplace.ca
5mo
@Tamasg @fastfinge ah, thanks
0
0
0
0
User avatar
Tamas G @Tamasg@mindly.social
5mo
@fastfinge yeah, that is surprising to me too! The benjamin thing is for sure odd. Almost like it's too squished and narrow to fit into the sample rate at that point.
1
0
0
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
5mo
@Tamasg Almost like the energy from the pulses is too narrow somehow? Maybe. Hmmm. If you change the sample rate on IOS, eloquence develops a weird lisp. So this tells us that whatever they're doing depends on the weird sample rate to sound good.
1
0
0
0
User avatar
Tamas G @Tamasg@mindly.social
5mo
@fastfinge I guess the good thing is the voice parameters themselves are very easy to tune in the NVDA driver itself, at least to change the numbers for that specific voice. So it wouldn't be hard to mess with that and find a new "person" through the driver that way, then it's very easy to use elsewhere. But the driver's not a bad playground for just tuning voice characteristics itself.
1
0
0
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
5mo
@Tamasg Yeah, that's what I'm fiddling with. Just made everything worse, though. haha.
1
0
0
0
User avatar
Tamas G @Tamasg@mindly.social
5mo
@fastfinge that's where AI might be useful, maybe. If it can understand the parameter relationships better between each and tune more carefully, who knows. I partially tried that yesterday with a new "reed" voice but it sounded too close to Adam so meh, reverted it. There's probably also more frame params to control as part of the voice, since they're tied to the 47 parameters you get when you enable the "exposeExtraParams" flag in the driver code. But we wouldn't want to make a voice that sounds like PC speaker tones either or something nuts like that, so it has to be carefully set. And if you look at the key for Caleb you can see they set some of the Aspiration stuff that say, Adam leeaves alone. Things like that.
1
0
0
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
5mo
@Tamasg Speak for yourself! This voice I made that just makes high pitched squealing noises that causes weird harmonics in my headphones is exactly what I wanted. Now I have an on demand headache generator. Perfect for work meetings I'd rather not attend! LOL JK
0
0
0
0