Reply to post

Hot!Support for SSML?

Author
woole123
User
  • Total Posts : 0
  • Reward points: 0
  • Joined: 2018/11/21 13:10:39
  • Status: offline
2018/11/22 01:46:04 (permalink)

Support for SSML?

Hiya new user here, I notice in the list of FAQ type threads, there is mentioned of SSML and it working at least partly, voice dependent? However I'm finding no real way to make it function with any voices I can find?
 
I've tried a few tags like..
 
<emphasis> tags, the voice reads the text, however the tag has no effect at all
<prosody> tags, I tried changing pitch via this (what I mainly wanted to try was pitch being alterable within the text rather than X value for everything) but the voice will not read the text within the tags any more. I tried both <prosody pitch="+X%"></prosody> and <prosody pitch="-/+Xst"> </prosody>.
 
The only tag which seems to do something is <break> where you can set say a 500ms delay.
 
I also notice if you type out any SSML tags they are invisible in the paragraph of text and when you then go to edit the text, much of the time you can't see them any more, but they do exist as almost hidden characters (if you delete all the text between where you placed them, the app seems to acknowledge there is some hidden remaining text as it shows a random ".." rather than what it looks like when you clear all text.
 
Just wondering if the lack of SSML support for these tags is more of an android issue? As I notice if I search SSML and android very little comes back.
 
cheers.

5 Replies Related Threads

    Admin
    Administrator
    • Total Posts : 275
    • Reward points: 0
    • Joined: 2010/11/22 00:00:00
    • Location: USA
    • Status: offline
    Re: Support for SSML? 2018/11/22 08:13:30 (permalink)
    Which SSML tags are supported and how depends only on the TTS voice you use. I cannot do anything else, except passing text to TTS engine, and including the tags as needed. The <break> is the most common SSML tag supported by different voices, but not all of them.
    Also, how do you enter the SSML tags. Through the "Edit text" function on the main menu? This may not work well, I would need test the issues here some more. The way I used them so far was with "Edit speech" feature on the Settings menu, e.g. if you want a 500 ms pause on each comma, do a replacement:
     
    Replacement kind: Case sensitive     (this does not matter much here, except that this is the fastest)
    Pattern: ,
    Replace: ,<break time="500ms"/>
     
    Greg
    woole123
    User
    • Total Posts : 0
    • Reward points: 0
    • Joined: 2018/11/21 13:10:39
    • Status: offline
    Re: Support for SSML? 2018/11/22 12:33:41 (permalink)
    ah I see fair enough, well I tried with a few voices I had, Samsung and Google's offerings, then one from CereProc and an apk I found for an old IVONA voice.
     
    Yeah I was doing edit text, tried edit speech the emphasis and prosody tags still have the same effect on the voices I have (emphasis does nothing, voices refuse to speak any text with prosody tags)
    Admin
    Administrator
    • Total Posts : 275
    • Reward points: 0
    • Joined: 2010/11/22 00:00:00
    • Location: USA
    • Status: offline
    Re: Support for SSML? 2018/11/22 15:25:06 (permalink)
    Yes, standard is one thing, the actual implementations - another. However the voices should still speak correctly with SSML tags embedded, even if they don't handle them. They would stop speaking only if the tag coding was incorrect. I maybe could point errors if you gave me concrete examples of what exactly you did and where.
    woole123
    User
    • Total Posts : 0
    • Reward points: 0
    • Joined: 2018/11/21 13:10:39
    • Status: offline
    Re: Support for SSML? 2018/11/23 12:37:12 (permalink)
    Yeah fair enough, I was trying simple text sentences. I was unsure in what format it would want a prosody pitch change, Amazon's TTS does percentages, Google does stages, so I tried both..
     
    So say with a Google voice I'd try...<prosody pitch="+20%">Hello there</prosody> or <prosody pitch="+2st">Hello there</prosody>. In both cases that would cause the voice to not speak those words, if I did normal words around it.
     
    Say. Hello there, <prosody pitch="+20%">how are you?</prosody>, that's good.
     
    The voice refuses to say any of this text. This is the same either if I write the SSML via edit text or if I use symbols and make edit speech transform it into those.
     
    AH! Just edited this, didn't think to test them but there are the voices which require internet connection. These do the tags correctly and weirdly enough without changing anything, now the offline ones at least try to pronounce the words within the tags though it has no effect, if the wifi is on...so odd.
     
    It might be no offline voices can handle them right, or at least none I've found.
     
    What happens with the online voices, do they send the text elsewhere then return it?
    post edited by woole123 - 2018/11/23 12:46:53
    Admin
    Administrator
    • Total Posts : 275
    • Reward points: 0
    • Joined: 2010/11/22 00:00:00
    • Location: USA
    • Status: offline
    Re: Support for SSML? 2018/11/29 13:54:54 (permalink)
    The online voices, currently only in Google TTS set I believe, send text of each sentence (actually, what @Voice highlights in yellow) to the Google server, and receive back "wave" data used to emit sound. How exactly they work, and which SSML tags they handle or not, I cannot say. They are products of Google company, not mine.
     
    I tested a sentence like:
     
    Just a test: <prosody pitch="+20%">Hello there</prosody>! test end.
     
    with Google local voice it said everything, but without any changes in pitch. However, entering tags like this through the "Edit text" function on @Voice menu does not work correctly, one must do such things through "Edit speech".
     
    Greg
    Jump to:
    © 2018 APG vNext Commercial Version 5.1