| VoiceXML 2.1 Development Guide | Home | Frameset Home |
| contour | Data Type: CDATA | Default: Optional |
| The contour attribute denotes the pitch contour for the text nested within the parent element. This attribute is non-functional on all TTS platforms, (Speechify and Rhetorical). | ||
| duration | Data Type: CDATA | Default: Optional |
| The duration element specifies the time , (in seconds or milliseconds) in which to read the enclosed text. | ||
| volume | Data Type: CDATA | Default: Optional |
| The volume attribute specifies the volume of speech output. | ||
| <?xml version="1.0" encoding="UTF-8"?>
<vxml version = "2.1"> <meta name="author" content="Matthew Henry"/> <meta name="copyright" content="2005 voxeo corporation"/> <meta name="maintainer" content="YOUR_EMAIL@HERE.com"/> <property name="nuance.core.tts.ResourceName" value="en-US.F019" /> <form id="F_1"> <block> <prompt> <prosody duration="8000ms"> Purple Haze, was in my brain. </prosody> <prosody duration="50ms"> Lately things dont seem the same. </prosody> <prosody rate="slow"> Acting funny, and I dont know why. </prosody> <prosody rate="fast"> Excuse me, while I kiss the sky! </prosody> <prosody volume="soft"> Purple Haze, all around </prosody> <prosody volume="loud"> Dont know if I am coming up or down. </prosody> <prosody pitch="low"> Am I happy, or in misery? </prosody> <prosody pitch="high"> Whatever it is, that girl put a spell on me! </prosody> </prompt> </block> </form> </vxml> |
| ANNOTATIONS: EXISTING POSTS |
mallu
|
|
| Prosody works with Rhetorical TTS Engine...
http://docs.voxeo.com/voicexml/2.0/frame.jsp?page=appendixn.htm |
|
jimjetjock
|
|
| Ah the many talents of Mr. Henry. Not just a technocrat, but a poet! | |
chakri.950
|
|
| Hi,
I would like to reduce the speed of the text for that I am using <prosody> tag. Problem is that I have tested in VOXEO server it worked fine, where as I test with my voice server I could not find the significance of <prosody> tag(its not working). The reason might be some TTS engines may ignore SSML tags. Could you please suggest what is the alternative that I can use to reduce the speed of the TTS. Thanks in Advance. KALYAN M |
|
MattHenry
|
|
|
Hello Kaylan, I am not clear on a few of the specifics regarding your post, so please bear with me: I have to make an assumption or two based on the context of your inquiry, so if my assumptions are off-base, please do clarify so that I can provide you with an accurate reply. I am going to assume that when you say your "voice server" this means a locally deployed instance of the Prophecy software, and that when you say you "tested on the voxeo server" that you mean a test application mapped to our developer portal. If this is reasonably accurate, then I should mention that you are correct: Not all TTS engines fully support SSML markup, but the default prophecy TTS engine should indeed support <prosody>, per [url=http://www.vxml.org/p8_tts.htm]our documentation link on TTS/SSML[/url]. If you are having problems with the implementation of this tag, it is suggested that you provide application code and debug logs to a private account ticket so that we can provide you with a solution that will work on the specific deployment that you are having problems with. ~Matthew Henry |
|
VoxeoDustin
|
|
| Hey Kaylan,
As a followup to Matt's post, I tested <prosody> using the rate attribute on Prophecy TTS, Rhetorical and Speechify and all of them worked as expected. I used to below code to test this: <?xml version="1.0" encoding="UTF-8"?> <vxml version = "2.1"> <form id="F1"> <block> <prompt> <prosody rate="x-fast"> This is extra fast </prosody> <prosody rate="fast"> This is fast </prosody> This is normal <prosody rate="slow"> This is slow </prosody> <prosody rate="x-slow"> This is extra slow </prosody> </prompt> <prompt xml:lang="en-us-fm3"> <prosody rate="x-fast"> This is extra fast </prosody> <prosody rate="fast"> This is fast </prosody> This is normal <prosody rate="slow"> This is slow </prosody> <prosody rate="x-slow"> This is extra slow </prosody> </prompt> <prompt xml:lang="en-us-fmj"> <prosody rate="x-fast"> This is extra fast </prosody> <prosody rate="fast"> This is fast </prosody> This is normal <prosody rate="slow"> This is slow </prosody> <prosody rate="x-slow"> This is extra slow </prosody> </prompt> </block> </form> </vxml> Let me know if you have any further difficulty. Regards, Dustin Hayre Customer Support Engineer II Voxeo Support |
|
punkin
|
|
| TTS engine reads fine for a short phase or if the sentence is short. Thus, I try to use <prosody> to slow down the speech. Unfortunately, it is difficult to use <prosody> to control the speech rate. Most of time, it doesn't sound natural to me if I use "slow" rate. This problem occurs to read a paragraph, especially the first few words and when a sentence is long (as if it had a very limited time to read all the sentences). In order to fix this, I have to physically break the sentence into a few portions, and each portion is separated by a comma. It could be problemic. How does the dialog know which paragraph or sentences are being read if the paragraph/sentences are passed by other application(s)? Nuance doesn't have to do this extra work. It reads as-is. I wonder if there is a better way to handle this.
Second, putting <break> inside <prosody> will receive silent and then exit. The debugger indicates it speaks but it doesn't. Is it a bug? It doesn't behave as it should be, in looking at the <prosody> spec here. |
|
VoxeoDustin
|
|
| Hello,
I'm not sure I understand the exact trouble you're having. All TTS engines will need some tweaking to achieve what sounds like natural speech. Even with Realspeak(Nuance) or Rhetorical engines, you will need to use punctuation and prosody to create pauses and intonation changes to create normal speech patterns. If there is something that is playing back incorrectly, I'd certainly like to address it. If you could give me an example paragraph that exhibits unexpected behavior, I'd be happy to take a look. A <break> within <prosody> causing TTS not to render is a known issue and has been corrected in Prophecy 9 and will likely be backported to Prophecy 8 on our staging platform. In the interim, you can either place the <break> outside of the <prosody> tag, or try your code on the Prophecy 9 development platform available within your Evolution account. Regards, Dustin Hayre Customer Support Engineer II Voxeo Support |
|
punkin
|
|
| Even reading a date value doesn't sound good to me. For example,
<prompt> Now is <value expr="new Date().toLocaleString()"/>. </prompt> I don't know how to describe this. Simple paragraph like the following is not natural to me: Many teens don't even see this behavior as wrong, according to the poll. Among those asked, 36 percent say that downloading a paper from the Internet was not a serious offense; 42 percent believe that copying text from Web sites was either a minor offense or not cheating at all. And 22 percent of those asked didn't feel that reading from notes on a cell phone during a test is cheating. Yes, if the data can be heard beforehand, we may be easily to tweak or fine tune the speech patterns. However, it the data is being passed by Web application real-time to the speech engine for interpretation, there is no way that we can do the tune-up. I guess it may be the quality issue of the TTS. When I loads this to my bevocal account, Nuance TTS reads fine to me. Sorry, there is no offense here. As far as I test, the TTS interpretation on Nuance sounds more natural to me and I don't need to do speech tuning or customization. But your support is fantastic. I wish that it could be improved in the future. I really like your service. |
|
VoxeoDustin
|
|
| Hello,
Below is some sample code using some of the Speechify and Realspeak(Nuance) voices on our staging environment. If you prefer one, you are welcome to use it for testing, however in a production environment, there is a surcharge for premium TTS voices. I assume the voice you are using with Bevocal is a Realspeak voice, so if you can determine which voice it is, we can certainly see about getting that setup for you on our hosted platform. <?xml version="1.0"?> <vxml version="2.1"> <form> <block> <prompt xml:lang="en-us"> Many teens don't even see this behavior as wrong, according to the poll. Among those asked, 36 percent say that downloading a paper from the Internet was not a serious offense; 42 percent believe that copying text from Web sites was either a minor offense or not cheating at all. And 22 percent of those asked didn't feel that reading from notes on a cell phone during a test is cheating. </prompt> <prompt xml:lang="en-us-fm3"> Many teens don't even see this behavior as wrong, according to the poll. Among those asked, 36 percent say that downloading a paper from the Internet was not a serious offense; 42 percent believe that copying text from Web sites was either a minor offense or not cheating at all. And 22 percent of those asked didn't feel that reading from notes on a cell phone during a test is cheating. </prompt> <prompt xml:lang="en-us-fmj"> Many teens don't even see this behavior as wrong, according to the poll. Among those asked, 36 percent say that downloading a paper from the Internet was not a serious offense; 42 percent believe that copying text from Web sites was either a minor offense or not cheating at all. And 22 percent of those asked didn't feel that reading from notes on a cell phone during a test is cheating. </prompt> <prompt xml:lang="en-gb-fm1"> Many teens don't even see this behavior as wrong, according to the poll. Among those asked, 36 percent say that downloading a paper from the Internet was not a serious offense; 42 percent believe that copying text from Web sites was either a minor offense or not cheating at all. And 22 percent of those asked didn't feel that reading from notes on a cell phone during a test is cheating. </prompt> </block> </form> </vxml> Regards, Dustin Hayre Customer Support Engineer II Voxeo Support |
|
xyliuCityU
|
|
| Hi,
I'm writing to ask whether Rhetorical TTS engine is still available now. If not, anyone has a Rhetorical TTS engine at hand? Now, I have Vocalizer 4.0 and RealSpeak 4.0 at hand. I want to make a performance comparison between Vocalizer 4.0, RealSpeak 4.0, and Rhetorical TTS engine. Thank you in advance! Regards, Cynthia |
|
voxeoJeffK
|
|
| Hello Cynthia,
It looks like you've opened a private account ticket with you inquiry as well. We'll focus on that ticket, so as not to create confusion. Regards, Jeff Kustermann Voxeo Support |
| login |