You are not logged in. Please login or register.
AudioGames.net Forum → Off-topic room → windows 10 higher quality voices for narrator, sapi 5 and windows one
Hello all hear is a feedback link I am sharing with you from the feedback hub. please upvoat this feedback in the feedback hub so we can see higher quality voices in narrator, sapi 5 and windows one cor voices come to life! https://aka.ms/AA1q5ve
+1
Link isn't working as of this writing, though will indeed vote up when its available.
Hello,
Are you running Windows 10? You have to be running that and have Feedback Hub signed into your Microsoft account in order to upvote this link.
Thanks,
Brandon
It didn't work earlier but now it does, so I upvoted it.
I'll definitely upvote this. The voices don't speak fast enough for my liking, the pauses for punctuation are far too long (IMO), and ... They just don't sound good to me in general.
keep your upvoats comming people? the feedback link is in post one! any who hasn't upvoated yet please click the link and upvoat now. those of you who are running windows 10 need to click this link. we need mor upvoats!
Hi,
To add to Mayson's comments about upvoting, you also have to consider though, what this change means. And this is honestly a hard one to give you an audio overview of. The only place you would hear this on Windows right now is when typing individual characters in SAPI5 with NVDA, or individual characters with Narrator.
Given this it may first appear that these are recordings. However, a certain product which I mention in various comments on this feedback item, Microsoft Learning Tools, uses this for everything, so obviously they can't record every single bit of text that one is likely to write.
With this being said I'm just trying to get this link out there so that we can see how people really feel about this. I originally posted this to the Windows 10 forum for screen reader users. I created this feedback about 4 months ago now and I started to think it beneficial to really spread the links. So if anyone's interested, I put out another link talking about something related to changing the Windows startup sound because the one we have is 11 years old now, but we can always have that on a separate thread.
I just thought this would serve as some helpful background so you have a better idea of what this TTS change could mean.
And @JasonSw, I don't know if this is the case, but given my tests in Learning Tools I couldn't pick up on long pauses in there, so maybe if this is changed in Windows it will fix that for you. I don't know, I'm just saying that could be a possibility.
Thanks again very much for providing your opinions on this as your upvotes are ultimately what determines whether anything happens. That's not just for this, though, it influences all aspects of things that they work on.
Thanks again,
Brandon
Hi,
Ah yes, I see your comments in there, too.
Thanks again,
Brandon
Reading is definitely unatural with non English Voices. Japanese in particular, at least when sed with NVDA seems to read all the puncutation unles you turn the punctuation level to none, though I don't know if that's an NVDA thing or the way the TTS itself handles it. But they definitely could sound better, though right now they're what I can live with...
Hi,
To address another concern that Jason was talking about, have you tried going into the Windows 10 Settings, under Time and Language, then Speech, and adjusting the rate slider? This will cause the Windows OneCore Voices driver in NVDA to follow suit and increase the speed accordingly.
Thanks,
Brandon
@7, Yeah, I believe the keyboard recordings are a specifically defined domain for the TTS engine. The data files seem to be in c:\Windows\Speech_OneCore\Engines\TTS\language code\NUSData. Most of the domains have .NU2, .RAD, .UNT, .WIH, and .WVE files, with the WVE files likely being the files that contain the recordings themselves, however I'm unable to extract them.
Hi,
Another thing though, is I'm wondering how the voice is this high quality for everything in their Learning Tools product. You can type a document and its text is spoken with this high quality.
Thanks,
Brandon
Probably because it's a cloud product. I believe the current voices are using small scale neural networks, and in the cloud they don't have to worry about them being CPU intensive.
Hi,
I thought about that, but thought I'd give it a shot.
Thanks,
Brandon
I don't recognize the female US English voice, though.
how do I use this learning tools thing? I can't demo the voice unless I know how. And there's also the Cortana wikipedia read-back, though I can't get it to read it back to me, it just opens it in bing on edge.
Type text and click try it out. You may need to press it twice for the actual playback controls to appear.
no I mean where do I get it, I've never heard of it.
Ok, after adjusting the slider in speech settings, the voices do speak much faster. Microsoft should really rethink that setup, though, as unless you know exactly where to look, you won't know how to make the voices speak faster, or that they can speak faster at all.
I still say that the pauses for punctuation such as commas and periods are too long, even at faster speeds. At the default speed, the pauses are ridiculously long, about 1-2 seconds between sentences, for example.
That speech rate thing is actually in the NVDA user manual, but yeah NV access did say that they've spoken with Microsoft about it and that it should be changed, but so far it was not.
Learning Tools is here. @JasonSW, while this is advanced, you can actually edit the pause length in the INI files for the voices. It makes them quite a lot faster and more pleasant to use. This shouldn't be necessary, though.
Pausing is a thing I wish synth manufacturers would not do, or at least give you control of, and I mean fairly easy to access control. I get it for things like train announcements and stuff of that nature, but when you're using a screen reader, its no bueno.
Yeah, and apparently Narrator uses an API to adjust the pauses, but it's private and would not be legal to use.
Hi,
This is correct. Narrator uses a private api to access additional settings. This is why the speech rate always goes faster in Narrator without having to adjust it in Settings.
This being said I wonder what Freedom Scientific does with JAWS and these voices? They can apparently speak really fast, too, just like Narrator, without going into speech settings in Windows Settings.
Thanks,
Brandon
AudioGames.net Forum → Off-topic room → windows 10 higher quality voices for narrator, sapi 5 and windows one
Generated in 0.023 seconds (61% PHP - 39% DB) with 11 queries