search

davidacm / NVDA-IBMTTS-Driver

This project is aimed at developing and maintaining the NVDA IBMTTS driver. IBMTTS is a synthesizer similar to Eloquence. Please send your ideas and contributions here!
GNU General Public License v2.0
60 stars 23 forks source link

What is the best version to be able to use IBMTTS with 22 khz, and where to get these libraries? #108

Closed titet11 closed 9 months ago

titet11 commented 9 months ago

@lbk2907

I am using the eci.dll libraries that come in the file you sent me but the truth is there is not much difference.

Actually all versions still have sound problems at 22 khz.

I thought it would be worth it but I'm getting disappointed.

titet11 commented 9 months ago

@rmcpantoja

I honestly don't think it's worth trying them, they sound horrible.

lbk2907 commented 9 months ago

Do you have any other version except the one that I send to you?

titet11 commented 9 months ago

The problem with 22 khz manifests itself when the letter S is pronounced

titet11 commented 9 months ago

@lbk2907

The truth is that I don't have other versions. but since I found several IBMeci.dll inside the file, I thought that replacing this library was more than enough, or I was wrong.

lbk2907 commented 9 months ago

Actually yes. But like you said, nothing different with the 22KhZ sound.

titet11 commented 9 months ago

@lbk2907

So does it mean that if you get another version, it will be enough to replace the IBMeci.dll library without making any further changes?

titet11 commented 9 months ago

@lbk2907

I think the version I just found is the same one you provided; the difference is that this file does not have other versions for the IBMeci.dll file.

archivos: IBMTTS 6.7.3.2 comprimido:

https://web.archive.org/web/20210720071828if_/http://grossgang.com/tts/old_software_synths/viavoice6732.7z

IBMTTS 6.7.3.2 en formato exe.

https://web.archive.org/web/20210720074324if_/http://grossgang.com/tts/old_software_synths/viavoice6732l.exe

lbk2907 commented 9 months ago

So does it mean that if you get another version, it will be enough to replace the IBMeci.dll library without making any further changes?

As far as I know is yes. Maybe it depend on version also.

titet11 commented 9 months ago

@Cleversn

Friend, I'm still waiting for the file, could you give me an estimate of when you could send it please?

lbk2907 commented 9 months ago

@Cleversn Later when you get the file, may you forward it to me? Thanks.

lbk2907 commented 9 months ago

@titet11

In terms of file size, you're looking at about 9 GB or so. And that's with both the 22 kHz and 8 kHz concatenative voices in the same package.

I think what @RetroBunn mean about the file is version 7.0.0.0. Yeah, the package is around 9GB++. But like what @Mohamed00 said.

The 22 kHz support was removed in 7.0.

The file that you ghave me just now is 3.85GB in size and has include the concatenative voices. All language in 1 package. Please take note that not all language support concatenative voices.

RetroBunn commented 9 months ago

Only two voices support 22 kHz, and they are both US English. If you set the variant to either Rocko or Grandma, you'll get new voices named Michael and Lisa, respectively, but their names won’t change. The rest of the voices I have are all 8 KHz concatenative ones for all supported languages. However, for some strange reason, some of the concatenative voices don’t speak immediately when you arrow through them. The only way I found to work around this is to go into Task Manager and kill ibmcmm.exe (IBM Concatinative Memory Manager), which is responsible for loading the concatenative voices into memory. And no, getting a higher RAM machine, like a 16 GB one, doesn’t help.

lbk2907 commented 9 months ago

Are you mean the version 7.0.0.0? Anyway, I will give it a try. Thanks.

RetroBunn commented 9 months ago

Are you mean the version 7.0.0.0? Anyway, I will give it a try. Thanks.

Yes, I’m referring to version 7.0.0.0.

lbk2907 commented 9 months ago

Surprisingly, after extract the file, the size is 14.2GB in total. Am I correct?

RetroBunn commented 9 months ago

Only two voices support 22 kHz, and they are both US English. If you set the variant to either Rocko or Grandma, you'll get new voices named Michael and Lisa, respectively, but their names won’t change. The rest of the voices I have are all 8 KHz concatenative ones for all supported languages. However, for some strange reason, some of the concatenative voices don’t speak immediately when you arrow through them. The only way I found to work around this is to go into Task Manager and kill ibmcmm.exe (IBM Concatinative Memory Manager), which is responsible for loading the concatenative voices into memory. And no, getting a higher RAM machine, like a 16 GB one, doesn’t help.

Something else I forgot to mention. If you end up killing ibmcmm.exe, be sure to switch to a different synth or go back to using the 11 kHz voices before doing that, then restarting NVDA before you can use concatinative voices again. In my experience, you only have to do this when switching from one concatinative voice to another, and you don’t hear any other voices speaking. Like I said, higher RAM machines don’t seem to help, at least in my experience, and anyone who’s used these voices feel free to chime in with an explanation of what’s going on.

titet11 commented 9 months ago

@RetroBunn

Hey... I asked you to send the file but at least I haven't received it. Could you please send it to me?

lbk2907 commented 9 months ago

Ok never mine. I receive your message.

RetroBunn commented 9 months ago

@RetroBunn

Hey... I asked you to send the file but at least I haven't received it. Could you please send it to me?

You haven’t received it because I’m not sending it anytime soon, and I will not be creating a burner email address just to send it (which, in turn, would give me plenty of storage).

Thank you for understanding.

titet11 commented 9 months ago

@RetroBunn

okey muchas gracias.