Page 1 of 1

Acronym too fast

Posted: Sun Oct 07, 2018 8:36 am
by hanan
Hello,
We're using "acronym" tag to have our payment confirmation code (alphabet and digits) read out to our customers. However, we've received complaints that it is too fast for them.

We've tried using slow and also rate percentage to slow it down, but that sounds very much unnatural and so would prove a poor use experience. We then tried delimiting the characters with commas, colons, periods, semicolons etc'

We then tried out adding spaces between each character of the confirmation code, which comes close to what we need, but now we receive complaints that "it sounds like the recording starts midway through the first character/digit".

Can you suggest any solution for this?

Re: Acronym too fast

Posted: Mon Oct 08, 2018 3:35 pm
by support
Hi,

Could you post a code example of your latest attempt? That will help us to see what kind of confirmation codes it is reading out. Also, which TTS engine and voice are you currently using?

Thanks,
Plum Support

Re: Acronym too fast

Posted: Mon Oct 15, 2018 2:04 am
by hanan
Hi,
Thanks for the response!

Here's some more info:
voice = "mike", American English (en_us) and we use AT&T Natural Voices.
An example (real production log) for this, currently with spaces as delimiters:

<?xml version=‘1.0’?><speak><speak xml:lang=“en_us”><prosody rate=“default”><voice name=“mike”><speak><say-as type=“acronym”>m z a d b x</say-as></speak></voice></prosody></speak></speak>

Re: Acronym too fast

Posted: Mon Oct 15, 2018 9:39 am
by support
Hi,

Unfortunately <prosody> can only get you so far. The problem is that it also slows down the pronunciation of the letters themselves. For example if you were to set <prosody rate="slow">M</prosody> the system would read it as ehhhhhhhmmmmm which does not solve the problem you need to address. We have found that the best solution for AT&T Natural Voices is a combination of small <prosody> changes and explicit <break> times gives the best overall result. Additionally you should use capital letters so that each letter is pronounced correctly. Here is the recommended code:

Code: Select all

<voice name="mike">
	<prosody rate="-25%">
		M<break time="500ms"/>
		Z<break time="500ms"/>
		A<break time="500ms"/>
		D<break time="500ms"/>
		B<break time="500ms"/>
		X<break time="500ms"/>
	</prosody>
</voice>
Let us know if you have any questions.

Regards,
Plum Support