Tips to manipulate the text to speach output

Forum|Forum|16 years ago
October 22, 2009
3 replies
4308 views

I am using the text to speech option in captivate 4

(Typing in slide notes and convert to speech)

I would like to make the output a little more “human”

Who has experience in manipulating the output?

I find out that inserting a “,” inserts a short delay.

Are there more ways to manipulate the speech?

Editing

This topic has been closed for replies.

S

sashby11

Participant

Try using a phonemes. Here's an example that provides excellent results with the Paul voice, for the word "baseline."

<vtml_phoneme alphabet="x-cmu" ph="B EY2 S L AY N">baseline</vtml_phoneme>

There are many different standards available, but the CMU phoneme from Carnegie Mellon University might be the easiest start. There's a dictionary here that generates the pronunciation code for many words:

http://www.speech.cs.cmu.edu/cgi-bin/cmudict?

Enter your word there and look it up. You can also check "show lexical stress" and it will give you a phoneme that applies appropriate stresses within the word (i.e. which syllable is emphasized). You can adjust that as needed. If the word you want isn't in the dictionary, lookup a similar word or part of the word to get started, then piece it together from there, adjusting the emphasis with "0", "1" and "2" for "no emphasis", "secondary emphasis" and "primary emphasis" as indicated on the CMU dictionary site.

The actual command "<vtml_phoneme..." comes from the neospeech manual someone else posted. http://www.neospeech.com/manual/vt_kor-Engine-API-References-v3.7.0%20(english_translation).pdf

SUMMARY:

put this line (below) in-line in your text (don't click "convert to speech" yet)
<vtml_phoneme alphabet="x-cmu" ph="B EY2 S L AY N">baseline</vtml_phoneme>
REPLACE the word "baseline" with your word, and
REPLACE the "B EY2 S L AY N" with the phoneme from the CMU dictionary
click "convert to speech".

I think you'll find it works pretty well. If you're interested in pursuing it further, check out the manual appendix C and look up some of the other phonetic languages, like x-SAMPA, which appears to be insanely detailed and powerful, but you might need a Ph.D. thesis before you can use it effectively.

A

Anonymous

Is there a way to emphasize a particular word using text to speech (rather than manipulating syllable stressing).

I want the voice to say "We will review...."

But it is saying "We will review...."

I tried using the phoneme VTML codes to adjust the lexical stress, but that did not accomplish what I wanted. Does anybody have any suggestions for to accomplish this?

Thanks!