[hts-users:04278] Re: generate speech with specified-duration of phonemes

[Dietmar Schabus <schabus@ftw.at>]

Hello,

take a look at the parameter -vp. If you use it, hts_engine will use the 
phoneme durations specified in the label file rather than the durations 
predicted by the duration model.

I have used this before to generate speech that has the same durations 
as recordings of the same utterances.

Best, Dietmar


On 2015-06-23 10:54, Tuan Dinh wrote:
> Dear all,
>
> I want to generate HMM-based synthesized speech with pre-specified
> durations of each phonemes. Is there any options of HTS-engine that can
> help me to do so?
> Thanks for spending your time on my question.
>
> Best regards,
> Dinh Anh Tuan


--
----------------------------------------------------------------------
DI Dr. Dietmar Schabus | Senior Researcher
phone +43 1 5052830-48 | fax -99 | schabus@ftw.at | 
http://userver.ftw.at/~schabus/

FTW Telecommunications Research Center Vienna
Donau-City-Straße 1/3 | 1220 Vienna | Austria | www.ftw.at