[DECtalk] introducing myself / synth singing

Talksina evvivajaws at libero.it
Wed Aug 6 06:52:07 EDT 2008


Hello.
I am an Italian blind girl, who loves to make synthesizers sing, I 
answered last week to a post about making the archive in .torrent 
file (I am interested about it)
As I wouldn't like someone to think I am such a lurker or some 
stranger coming to ruin the list's atmosphere...I love to make 
synthesizers sing. DecTalk first of all, and also the Italian one, 
but unfortunately that one sings just my language, there's no way to 
make it sing in other languages, as in dectalk there is not ability 
to make it sing italian!
well, to introduce myself in the audio way,
I am the author of this:
http://www.bluegrasspals.com/dectalk/Miscellaneous/blowing.mp3
or this
http://www.bluegrasspals.com/dectalk/Miscellaneous/crash_test_dummies.mp3
or this
http://www.bluegrasspals.com/dectalk/Miscellaneous/titanic_soundtrack_by_Jois_from_italy.mp3
well, Jois (pronounced as Joys) was my old nickname. Very old, but 
now I am Talksina in every Internet community.
It would be nice if those files could be placed in a directory called 
Talksina, on the archive, but, tell me if you need me to put the ID3 
tags and send them back, to make it easier I could put all my DecTalk 
productions on a zip file, then send it by itself or downloadable via 
link or torrent. But, as I am leaving for holiday  on Saturday, 
coming back the 1st of September, we'll talk about it in that date.
Well. Making synthesizers sing, has always been my passion since 
2000, and started playing with DecTalk.
But unfortunately, I found no Italian-speaking singing synth. As 
English is not my mother language, the effort for me is double as I 
must, first of all hear the song's words, then, get the lyrics from 
Internet and of course the music in Midi or karaoke version. Then, 
try to find the right pronunciation with phonemes, sometimes I make 
horrible mistakes, as I am terribly influenced by my Italian accent, 
such as writing the word "windows" in phonetic alphabet instead of 
"wixndowz" I wrote "wiyndows". ARGH! So, if I get a non-English 
pronunciation with English voice,the sound is rediculous. So, the 
right way I found, is to make DecTalk read the word as it is, and 
find out the phonemes that match the correct pronunciation.
That's why I am making it sing lesser and lesser, I need such an 
amount of time!
In 2003 I discovered that my licensed Italian Loquendo robotic male 
and female voices could sing, I was just playing with a TTS, made a 
mistake and the 2 words just sounded with different music notes. It 
looked really as it was singing. And that was!
So I started enjoying myself making songs and tests
and, the best thing of this synth is that it doesn't need to convert 
the audio file into mp3 or wav to try if the sound is good, if there 
are mistakes, it can directly speak, it's sapi5 compliant and works 
with any TTS-driven software that support synth-specific commands.
a sample (italian song) here, just to make you hear what I am talking about.
http://www.audiomontaggi.net/erfinestra/cremonini_demo.mp3
The trouble of this synth is that unlike DecTalk, it has no specific 
adjustable parameters to make it sing, you adjust silence with 
punctuation marks and writing a letter as many  times as you wish the 
sound to be long. So, counting that a letter is about 200 
milliseconds, you must calculate how many milliseconds are needed to 
match whatever note in a music file, to syncronize voice and music. 
And if nothing fits, you must play with voice rate parameter
OK... bye ciao.




More information about the Dectalk mailing list