Attachment nvda work to fast added by Twinsen on 2012-06-12 13:48
Description:
In this log you can see how NVDA work to fast

Comment 1 by jteh on 2012-06-12 22:56
Please explain how you define "too fast" and "too slow". What do you mean? Relative to what? Have you tried adjusting your speech rate? What synthesiser are you using?

Comment 2 by Twinsen (in reply to comment 1) on 2012-06-13 15:12
HI! I mean speed reading/talk of the text.I don't adjust the synthesiser.I use standart synthesiser e-speak.

Please explain how you define "too fast" and "too slow". What do you mean? Relative to what? Have you tried adjusting your speech rate? What synthesiser are you using?

Comment 3 by jteh on 2012-06-13 17:17
I still don't understand the issue. If NVDA is speaking too fast or too slow, you can adjust the speech rate. Or are you saying the speech rate changes randomly?

Comment 4 by Twinsen (in reply to comment 3) on 2012-06-14 15:15
Replying to jteh:

I still don't understand the issue. If NVDA is speaking too fast or too slow, you can adjust the speech rate. Or are you saying the speech rate changes randomly?

Ok! tell me why nvda changes speed rate without me? I don't press (shift+ctrl+right arrow) ?

Comment 5 by briang1 on 2012-06-14 17:50
You mention Russian then you mention Greek, could it be that the synth produces these at apparently different rates. I suspect few of us understand enough of either to understand the problem here.
I do know some web pages do switch the language, but I suppose if the page itself is very busy it might slow the delivery down, but I've never heard it speed it up!
Are there other pages where this occurs?

Comment 6 by MHameed on 2012-06-21 05:12
Hi,

This is a bug with espeak. There are some characters that causes it to change its reading speed, and it looks like you found another of these symbols.
before symbols such as ++ -- __ ^^ @@ %% causes speedups/slowdowns.

Which espeak language do you use when this happens?

Comment 7 by Twinsen (in reply to comment 5) on 2012-06-27 19:22
Hi, i don't remember another sites.Replying to briang1:

You mention Russian then you mention Greek, could it be that the synth produces these at apparently different rates. I suspect few of us understand enough of either to understand the problem here.

I do know some web pages do switch the language, but I suppose if the page itself is very busy it might slow the delivery down, but I've never heard it speed it up!

Are there other pages where this occurs?

Comment 8 by Twinsen (in reply to comment 6) on 2012-06-27 19:29
I use the Russian languageReplying to MHameed:

Hi,

This is a bug with espeak. There are some characters that causes it to change its reading speed, and it looks like you found another of these symbols.

before symbols such as ++ -- __ ^^ @@ %% causes speedups/slowdowns.

Which espeak language do you use when this happens?

Comment 9 by MHameed on 2012-09-22 11:34
@twinsen Sorry for the long delay, I didnt notice you replied.

Can you do the following:

1. find a webpage that has this problem.
2. copy and paste into wordpad the line where the problem first starts, i mean the line when you can hear it become very fast or very slow.
   Maybe you need one line for fast, and another line for slow.
3. now in wordpad, if you go up and down, do you hear the same problem?
4. save and attach the wordpad file to this ticket.

Thanks.

Attachment Error occurs.rtf added by Twinsen on 2012-09-24 19:54
Description:
The error ocur only if string started :PRO/Wireless : placed on new line.

Comment 10 by Twinsen (in reply to comment 9) on 2012-09-24 20:37
Replying to MHameed:

@twinsen Sorry for the long delay, I didnt notice you replied.

Can you do the following:

1. find a webpage that has this problem.
2. copy and paste into wordpad the line where the problem first starts, i mean the line when you can hear it become very fast or very slow.

Maybe you need one line for fast, and another line for slow.

1. now in wordpad, if you go up and down, do you hear the same problem?
2. save and attach the wordpad file to this ticket.

Thanks.

Ok! I attach Word pad document. Use Num Pad <+> to read this text.On this example you see how e-speak start work to slow. Tell me please how I can use controls placed in object ?I know button switch betwin objects but how use the objects ?

Using:

src/espeak-ng -vru "Hello Александр Македонский, how are you?"

Both English parts are spoken normally, while the Russian appears sped up. As such, I think it is that the Russian phonemes have a shorter duration than the English ones, so appear faster at the same speech rate, although I have not confirmed this.

I assume you are one of the developers of Espeak, since nobody has said, but I've often wondered how the samples are produced, as we consistently hear that the us English is not very American sounding, and listening to it, it seems to be lacking the twang extension one hears in us voices based on eloquence etc. Not that I like that voice, nor am I an American :-)

The Quincy voice in nvda is one I fiddled with and made, and many seem to like it.
So how exactly are the samples made?
As this is a ticket, you might wish to transfer this to a list instead.

bglists@blueyonder.co.uk
Sent via blueyonder.
Please address personal email to:-
briang1@blueyonder.co.uk, putting 'Brian Gaff'
in the display name field.
----- Original Message -----
From: Reece H. Dunn
To: nvaccess/nvda
Sent: Sunday, January 10, 2016 9:56 AM
Subject: Re: [nvda] NVDA read the web pages to fast or to slow (#2441)

Using:

src/espeak-ng -vru "Hello Александр Македонский, how are you?"
Both English parts are spoken normally, while the Russian appears sped up. As such, I think it is that the Russian phonemes have a shorter duration than the English ones, so appear faster at the same speech rate, although I have not confirmed this.

—
Reply to this email directly or view it on GitHub.

Jonathan Duddington is the developer of espeak. I have been maintaining a POSIX/Linux build and the Android port for some time now. Recently, I have taken on the role of developing the espeak-ng fork.

Regarding how the samples are produced, the data is a mix of wave audio, spectral formant data and klatt formant data (for the klatt voices). These are selected and mixed according to phoneme rules. I don't currently fully understand how that code works, so I don't currently know how to modify a voice to produce the effect you describe.

@rhdunn, since this sounds like an Espeak issue, can you file an issue on this topic on the ESpeak Issues on Github and post a link to the issue here so that we can track it?

Looks like this issue has been resolved in Espeak. Closing