[prev in list] [next in list] [prev in thread] [next in thread] 

List:       festvox-talk
Subject:    Re: Problem in built speech output
From:       Alan W Black <awb () cs ! cmu ! edu>
Date:       2010-03-01 15:35:25
Message-ID: 4B8BDEBD.4030209 () cs ! cmu ! edu
[Download RAW message or body]

Thilini Liyanagama wrote:
> Hi,
> I'm testing the diphones which were manually labeled. It speaks out
> wrong diphones in different positions in the same wave file. But the
> diphone index entries have the correct start, mid and end point
> values.

Are all the waveform files of the same sample rate?

Can you tell if the actual times in your indexes are correct with 
respect to the waveform file.  If you can't find a working display 
method, you could use ch_wave to extract a diphone with those times and 
ensure it is correct.

Is it getting some other part of the same waveform file or a different 
waveform file?  (You check this with the ungrouped diphone voice and 
check to see which files are accessed (with ls -alutr).

Alan

> 
> Thanks and Regards,
> Thilini
> 
> 
> On Mon, Mar 1, 2010 at 1:52 PM, deneth karunarathne <deneth9@gmail.com> wrote:
>> check the lexicon file
>>
>> On Mon, Mar 1, 2010 at 1:15 PM, Thilini Liyanagama <thilini85@gmail.com>
>> wrote:
>>> Hi Nickolay,
>>> I have built a new Sinhala voice.Then I used the SayPhones command and
>>> it gives correct sounds for diphones; 'silence + vowel',' vowel +
>>> silence', 'silence + consonant' and 'consonant + silence'. But gives
>>> wrong sounds for diphones in the form 'vowel + consonant'. The entries
>>> in the diphone index file give correct values for start point, mid
>>> point and the end point. I can't figure out the problem. What could be
>>> the issue? Please help....
>>> Thank you,
>>>
>>> Regards,
>>> Thilini
>>
> 

[prev in list] [next in list] [prev in thread] [next in thread]