Spoken Emphasis Mark-Up Language

[Graevarg]

Hey Gary,

I have become something of a VA addict here if late, the program is just so damn terrific! I am looking to get a pro TTS voice from INOVA ("Emma") and have been doing research on what I will call, for lack of a formal term I can find, "spoken emphasis markup language" (lol - SEML) - the ability to format text with additional non-pronounced characters that affect how contained words are pronounced. (The verbal equivalent of boldfacing, I suppose). So I wanted to go to the source -- does VA have, or do you know of, anything like that?

Case in point.  "What did you think was gonna happen?" spoken flat is a simple question.  "What did you THINK was gonna happen?" becomes very rhetorical / snarky. Sorry if I am being overly simplistic...!

Anyway, many thanks!

[Gary]

Hi, there.

You should be able to use SSML (Speech Synthesis Markup Language) with VoiceAttack.  Just pop all the SSML into the TTS box and you *should* be good to go.  I haven't documented it, because I haven't fully tested it or set a way to force ssml handling to be on/off.  The way it stands right now is kind of weird...  VA is looking for "<speak version=" (no quotes) in the tts text to turn on SSML handling.  I'll need to add an option to turn on ssml, as the version attribute is sometimes not located right after the speak element.

<speak version="1.0"  xmlns="http://www.w3.org/2001/10/synthesis"  xml:lang="en-US">

  What did you<emphasis>think</emphasis>was going to happen?

</speak>

Normally, this should work as you describe... that is, if the TTS voice supports the emphasis tag.  As documented by Microsoft, the Microsoft voices do not support it (also tried with Cereproc Heather and it doesn't work there, either).  Could work with IVONA, but I have no idea.  One day I'll swing back around and finish what was started with SSML in VA.

Hope that helps!

Gary

[Graevarg]

Sounds good, I will give it a try with the IVONA voice and see what I get. I may drop the SSML question directly to IVONA, if they respond I'll come back here with the answer.

[Graevarg]

well that didn't take but a second or two.  At least preliminary answer, according to the IVONA website, is YES, they support SSML. FAQ page with a fair bit of detail is here:  http://developer.ivona.com/en/ttsresources/ssml/ssml.html

[Graevarg]

OK, success! I got the IVONA voice "Emma". Installed easily and instantly worked with VA. So I then tried SSML and it worked in the Speech app that came from IVONA. So.... now I just need to test the same block of test with and without a strong emphasis on a test word to see if it carries through. Fingers crossed - will let you know!!

[Slannesh Scott]

I tried this using the IVONA2 - Amy voice, but in both preview and proper activation it goes back to the default Microsoft voice.

All my other TTS not attempting to use SSML works fine using the Amy voice.

Am I missing something?

[Gary]

It's not fully implemented, Slan.  I don't even think it's documented anywhere.  What I found is if my SSML was messed up while using Cereproc, the default MS voice was used.

Sorry about that... I'll move it up the chain to get it sorted out ;)

Gary

[Gary]

Hi, Slan.

I went ahead and put in a checkbox to process as SSML (so that SSML can have whatever attributes).  What I found after that was strange.  If I had invalid SSML, the TTS wouldn't play at all.  If the SSML *was* valid, and MS Hazel was selected, MS Hazel would not speak... it would defer to another voice.  In my case it was Cereproc Heather (installed for testing).  If I uninstalled Cereproc Heather, the TTS would defer to MS David.  If I selected Heather, David or Zira, TTS would work just fine with SSML.  If I removed the SSML and selected Hazel, Hazel seems to work.  I'll put the new stuff out there in the next release.

Gary 

On Wednesday, December 16, 2015 at 7:27:37 PM UTC-6, Slannesh Scott wrote:

[Slannesh Scott]

All seems a bit confusing switching between voices though! But it would be good to play with it in the next version :)

[brittan...@gmail.com]

Just a note for anyone who's having trouble with this - I have the Ivona AMY voice, and couldn't get it to work until I changed the xml:lang to en-UK (instead of en-US which a lot of examples use).  I guess the MS speech engine needs or wants a specific match to region/locale.  "en" by itself might work too.

The following works when Ivona-Amy is my default voice:

<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" 

    xml:lang="en-UK">

<p>

  That is a <emphasis> big </emphasis> car!

  That is a <emphasis level="strong"><prosody rate="x-slow" volume="x-loud" > huge </prosody> </emphasis>

  bank account!

</p>

</speak>

Cheers!

[Jorge S]

I mostly use Ivona Brian voice (very Jarvis like voice :D ), with your change it now works perfectly, thank you Brittany :)

[Slannesh Scott]

Thanks Brittany, this was the bit I was missing :)

[Graevarg]

Brilliant Brittany, thank you!!
:)