[Parley-devel] Re: A pair of ideas...

[Massimo Lo Iacono]

On Thu, 24 Feb 2011 17:48:52 +0000
alsaf <alfraealba at gmail.com> wrote:

Hi alsaf,

I have doubts about using synthesized speak sounds too.

For me in the first place there is a physical reason in the sense that normally the quality of synthesized sounds is bad, and in the second place a more psychological reason because the sound of human voice, this is my opinion, isn't only a physical matter but it is also a manifestation of human spirit. I think an adult person can use this tool with sufficient consciousness but a child maybe not. I think in our world there are yet enough opportunities of alienation for young people with television, play station and other similar things. These are my doubts. 

> Maybe if the sound option could be linked into the IPA/Phonetic field of 
> each word? This would be done by taking the words IPA/phonetic field and 
> mapping each IPA symbol to an audio/synthesized clip and then outputting it?

If I understand right, I think the technical problem isn't at level of IPA/phonetic code, that is I think the bad quality problem is not resolved using "IPA/phonetic field and mapping each IPA symbol to an audio/synthesized clip and then outputting it". A program like Espeak produce yet a code like IPA from a text, according to vocabularies for different languages and these vocabularies would be prepared by expert people according to pronunciation rule for each language. So this code is different for each language. In a second step this code is used by the synthesizer machine (like Mbrola) to produce sound. The lack of quality is mainly in this second phase. For example I proved the "mb-de3" German voice vocabulary for Mbrola machine and it seem better than other.

At present I think to implement this functionality in Parley for myself and prove it to see how it is.

Good night
Massimo