Microsoft TTS working but others not reliably. Windows10

[Olias]

I couldn't find a post relating to this specific circumstances so if I missed it please point me to the correct link.

Background
VA has been on my system for two years. In fact, it's migrated through 3 different upgrades with no issues to report outside of my own learning curve. However, one nagging issue I have never been able to resolve is the inconsistent replay of non Microsoft TTS engines. I've only tested NeoSpeech VW Paul VW Kate and they work but only on the first initiation of the TTS response. They often do not work when I am building a command and using the "Say Something with Text-To-Speech" function's Preview button. If I save the command anyway and forego the preview, the command will always work but the TTS will only work once. Occasionally it will function again after a few minutes but I have yet to nail down under what circumstance it works and which ones it will not.

If I had to theorize, it's almost like the MS Text to speech engine is either hanging or crashing. However, I am not able to find any reference to T2S troubleshooting on the MS website, not even in the developers areas. This technology is almost like a black box to Windows and since the core speech engine hasn't been updated since 2011, it's starting to look like MS is abandoning this technology for perhaps Cortana which would really suck.  I also couldn't find any location on my system where diagnostic log was located or at least one that indicated failures of playback. If there is one, please let me know. I even enabled the "Speech-TTS" event log viewer and so far there haven't been any events written to the log.

How to re-create - (All testing was done with Plugins disabled)
Create any "Say something function"

• When one selects a voice other than the default MS TTS speech voices, it works the first time when clicking 'preview'
• If one selects any other non MS TTS voice, the other voice will not play. That happens 100% of the time.
• After Previewing the second non MS TTS, switch back to an MS TTS and those will work fine 100% of the time.
• Toggling back and forth between MS TTS and Non MS TTS worked in preview mode but only for the first non MS voice that was tested. The order of the first non MS voice tested did not matter but whichever one was tested last failed 100% of the time.

Other symptoms are that after a few minutes, the voice engines will somehow recover and work for a single instance, only to fail on the repeat invocation.

This error has been very consistent since I first installed VA three systems ago. It's not up until now that I've spent the time trying to diagnose what the problem is. My motivation now is twofold, I want to install other voices which are considerable expensive but do not wish to do so if they won't work with VA. Second motivation is to use different voices for different messages. I do not want to use the Microsoft Voices as they are harsh on the ears.

I've tried to use different options such as "complete command", "Use single TTS", "This completes all other.." and so on but this does not work. Restarting VA was the only way to get the commands to work. "Reset Active Profile" worked sometimes.


Note in these examples, I use a play sound event simply to give some kind of feedback that the command was executed. Disabling that did not affect the performance though.

[Pfeil]

If I had to theorize, it's almost like the MS Text to speech engine is either hanging or crashing. However, I am not able to find any reference to T2S troubleshooting on the MS website, not even in the developers areas. This technology is almost like a black box to Windows and since the core speech engine hasn't been updated since 2011, it's starting to look like MS is abandoning this technology for perhaps Cortana which would really suck.  I also couldn't find any location on my system where diagnostic log was located or at least one that indicated failures of playback. If there is one, please let me know. I even enabled the "Speech-TTS" event log viewer and so far there haven't been any events written to the log.

Just to confirm, it's the non-Microsoft voices that aren't working, correct? Those third-party voices would not normally use the Microsoft text-to-speech engine, but rather their own (proprietary) engine, controlled through the SAPI standard.


As you mention the issue occurs with voices from one specific vendor, it's worth checking whether it occurs with voices from a different vendor.

Ivona products (this is not intended as an endorsement, but they are known to work correctly in this context) have a trial version available (E.G. from this reseller; Other resellers and voice brands also available)

[Olias]

As always, thank you for the speedy response.

Yes, you are correct. These are SAPI5 products originally made by Neospeech AKA Voiceware now owned by readSpeaker which is now a services model based business and no longer supports a desktop end-user product.

I've looked at the Ivona products but at $45 a copy, one can see that this quickly becomes prohibitive for the objective I have which is to have multiple voices for commands. I supposed if one was only requiring one voice, then the price isn't too steep.

This is how it looks from the windows registry.


Here's the information on one specific voice. Note the lack of copyright notice. I couldn't find anything specific on the web to verify that it is indeed a SAPI5 component.


The only indication I have that it is SAPI5 and not SAPI4 is from Balobolka open source TTS player. I include this data only to show some of the forensics I came up with when digging into these two voices. If I could somehow make them work, then purchasing a third voice from a third party like Ivona or Cepestral isn't so out of the question. I may still give an Ivona voice a shot just to validate that it does in fact work.

[Pfeil]

I may still give an Ivona voice a shot just to validate that it does in fact work.

That was the intent, to rule out that this is an issue specific to "NeoSpeech" voices.