My report to NVDA

Integrating the Cepstral engine

My report to NVDA

Postby mikebayus » Thu Sep 25, 2014 12:38 pm

Hi all,
Yesterday, I decided to report an issue concerning Cepstral voices to NV Access, the developers of the NVDA screen reader for the blind.
The voice I use is Amy, but the issuewith NVDA and Cepstral, I believe, is not Amy specific. I have never tested this using any of the other Cepstral voices, but I think that the NVDA developers have been aware of this issue for quite some time, as it has persisted, as far as Cepstral is concerned, throughout NVDA's development. If the issue is Amy specific, then I would suggest that it might be "RIP Amy." I know that Amy has never been a favorite of Cepstral, and that she is considered old technology and is no longer in development. I don't really know to what extent the woman whos voice was used to build Amy is involved with Cepstral, and whether she would even be interested in re-recording her voice to make a larger data base in order to update Amy, but I do know that there are plenty of voices out there in the population of Pittsburgh that are youthful and sweet, who would be interested in such a project as building a Teenaged or young student voice for Cepstral, if they only knew about it.
Here is what I told NVDA, and the comments that were made by NVDA developers.
Opened 26 hours ago
Closed 17 hours ago
#4490 closed defect (cantfix)
Cepstral voices crash when adjusting the rate too fast
Reported by:
Owned by:


Operating system:
Windows 7
Blocked by:


Changes document entry (for developers):

When in the voice settings menu, and a Cepstral voice is picked, If tabing to rate, and pressing the up and down arrows too quickly or holding one or the other down to speed up or slow down the voice, The voice crashes, NVDA reinitializes, and if another voice is set as the default voice, the Screen reader reverts back to that voice with all settings, pitch, rate etc. set to there original settings, and if the Cepstral voice is the default, NVDA relaunches with the voice, but with it's original settings. If pressing the up or down arrow one increment at a time, and stopping until the new value is completely spoken, it works as it should. I have told Cepstral about this, and they don't know or care enough about NVDA to do anything about it. I have CereProc, Ivona, and Neospeech voices, and they work as they should.
Change History (3)
Changed 26 hours ago by mikebayus
? Cc added
Changed 21 hours ago by briang1
Well you have in a way answered your own question. Its their code that is the issue.
I'm not sure what could be done about this at the nvda side.
Changed 17 hours ago by jteh
? Keywords crash added
? Resolution set to cantfix
? Status changed from new to closed
Sorry, but there's no way we can compensate for bugs like this in specific synths. Any workaround we try to implement here is likely to disadvantage other synths. This needs to be fixed in the synth.

Well that's it.
I know it's a minor probllem, or so it might seem, but it does effect the reading of documents by blind people who want to read a line or a word at a time, or to adjust the rate of speech on the fly.
Cepstral is made in america, and I would think that it's developers would want it to work as perfectly as they know it should.
Posts: 21
Joined: Sat Jan 26, 2013 9:39 am
Location: Sarasota Florida

Re: My report to NVDA

Postby mikebayus » Mon Feb 02, 2015 11:47 am

Here is an update.
I have just installed a development snapshot for NVDA version 2015.1 as NVDA has updated it's version of python. I tryed replicating the behavior that causes my Amy voice to crash, and while she crashed the first time, she never crashed again even though I tryed replicating the crash many times.
Here is how to cause Cepstral to crash NVDA: (I don't think that it is just Amy, but you might want to install the Amy voice just in case.)
1. Make sure that you have NVDA version 2015.1 installed, and that the Cepstral voice is picked.
2. Go to the voice settings menu by pressing and holding the inseart key and while holding it down, press the letter N to get to the NVDA main menu. Then tab to voice settings and press Enter.
3. Tab to the rate setting.
4. Press the up arrow very fast until NVDA crashes. If it doesn't, press the down arrow after the skale gets to 100 Amy should crash, and NVDA should innitialize.
In the latest install of NVDA, it only happened once for me.
As you read in my NVDA support ticket, NVDA thinks that it something in Cepstral's code and not the fault of NVDA. They say to try to implement a work around would cause other voice synthesizers to malfunction.
With new NVDA, Amy only crashes once, and so far, I had to make that happen.
Posts: 21
Joined: Sat Jan 26, 2013 9:39 am
Location: Sarasota Florida

Return to Asterisk

Who is online

Users browsing this forum: No registered users and 1 guest