espeak-ng
espeak-ng copied to clipboard
Tamil / kannada / telugu: gemination instead of literally saying the consonant twice?
So, I was wondering, in the languages I mentioned in the issue title, instead of literally saying the consonant twice, shouldn't they be geminated instead? I hear this in other speech synthesizers speaking those languages and this was also done in Turkish at some point in this continuation of espeak. So, if, in fact, gemination is better, I don't know how to realize it and any help would be most appreciated. Thanks in advance!
Please give some example words and the output you expect.
An example in tamil would be the number 8. The two t letters in that are said twice instead of being geminated,which I think should be geminated.
Take a look at the file dictsource/ta_list.
See if you can find a pattern. For example, here's a sample of "88". It looks like the "eighty" sounds like you expect, but the "eight" doesn't. Is that what you mean?
$ espeak-ng -v ta 88 -X
Found: '_8X' [;en.bVttU]
Found: '_8' [;et.t.U]
;'en.bVtt,U;et.t.U
To me it looks the problem is in the use of the period "." used in the list. I can't find an explanation for it in the documentation, and few other languages use it.
So, read through the file and test the words that have a period between two consonants. Do they all sound broken to you?