@Richard.

Doh, I actually knew that about the English/French Canadian sleep.

Anyway yeah, I'm hoping some other members will come forward with what they are using and post some details like I did in post #71. @Rich. If your reading this, what OS are you using Cereproc with?

The only way that I see a change coming is for EZ-Robot to change their speech engine from the one that is supplied with windows to another one that is able to convert these voices. I dont know if that is a possibility. It would require quite a bit of work I would suspect and would require the purchase of a speech SDK.

I am still looking for a way to get the microsofts speech engine back to normal after installing the voices that we have.

@d.cochran.

Thanks David. So in your eyes, what do you reckon the problem actually is? I not that experienced in such things but I wouldn't have thought the guys changing their speech engine would solve it, as it's working for some people and not others using the same voices, ARC version, operating system, and in some cases the same make of computer. If it didn't work for everyone then I could kinda understand it.

But I could be wrong of course.

I am using the same development platform that I believe they are using.

The issue is that this platform works with TTS voices, which the ones installed are not TTS voices. This causes them to have to go through a separate driver to be converted to TTS. This isn't working with the current version of the objects that are available to the current development platform.

The reason that these work with some apps on your system and not others is because these are converted using a custom driver in that software that is able to convert these to TTS. This happens at runtime for the application.

The reason that it works on some computers and not others could be a couple of things. 1. The version of the controls that are being used are not the same, 2. There is something catching this conversion and handling the conversion to TTS on their system that is not on ours, or 3 They are using different voices that are true TTS voices.

A different SDK with a list of supported voices would eliminate this issue. The problem is that EZ-Robot would have to check and approve voices as they came out. I dont think this is the best use of their time by any means.

I am trying a voice from the company that Rich has used to see if that works. I will let you know what I find out. It may be that they use a different "driver" that will catch these and convert them better.

The only voices supported natively by the speech synthesizer that is probably being used by EZ-Robot on Windows 8.1 are the native Windows voices. I am not surprised by this at all as Microsoft would have to verify and validate voices as they were being produced. I personally think that there should be some sort of voice verification group at Microsoft if they ever want their speech engine to really take off, but who am I? This would allow voice vendors to put something on their voices saying "Microsoft Approved" or something like that. I will find the version of the DLL that is on my system that is used for the speech synthesizer. It may help to point to a possible solution.

@d.cochran,

It seems from your results that Microsoft has probably broken the SAPI API in 8.1. These 3rd party voices all follow the MS SAPI 5.x standard. (note: I have the problem with the broken MS voices on a machine that I did not attempt to install Cepstral on. Just MS, no 3rd party voices).

I am not actually surprised if Microsoft broke SAPI (unintentionally, but through neglect). I have some inside information about how they started to move in a different direction for speech, abandoned that, and now are going in yet another direction with Cortona. SAPI has not had any Microsoft love for a while.

(edit: I was typing this before you entered post #93. leaving in place as I still think this is the issue, but it seems more intentional, with the current SAPI requiring "signed" voices, and perhaps the 3rd parties have not done that).

Alan

I suspect the version of system.speech.dll that they are using is from .net framework 4.5.0 and is product version 4.0.30319.17929.

.net Framework 4.5.1 has version 4.0.30319.18402. This is the version that I am testing with but would be happy to setup a .net 4.5 project and test the other one. I might setup a project that uses a later version of .net framework and see if it makes a difference. I doubt that it will.

I dont know if any of the .net framework service packs have changed this DLL but I suspect that they have. It is possible that those who have these versions working have a different version of this DLL.

This dll can be found in C:\Program Files (x86)\Reference Assemblies\Microsoft\Framework.NETFramework\v4.5 or whatever version you want to look in on your machine. I believe that ARC is compiled to be 32 bit so this should be the correct location.

@David.

Thanks for your insight my friend. The voice that Rich uses is from Cereproc which I have a 60 day trial period for, and that's not working for me in ARC either. It would be interesting to see what results you have with it. Still another curiosity is that a lot of us can only use one of the three native Microsoft voices with Win 8.1.

I am looking to see if the Speech.Speech.xml file was messed up by the install. It may also be buried in the registry. One thing is for sure, there is not much information on the internet that offers any help.