[hts-users:03136] Re: Reg. Integration of HTS voice

[Keiichiro Oura <uratec@xxxxxxxxxxxxxxx>]

Hi,

I think 6 secs are too long for the dialogue system.
It should be within 0.1 sec.

1. Program
If you use HMGenS in HTS & mglsadf in SPTK for synthesis, you should
use ``Festival2.1 and hts_engine API'' or ``Flite+hts_engine.''

2. Audio
If you play the audio file after synthesis process is completed, the
synthesized samples should be send to the audio devices directly in
synthesis process.
hts_engine API support ``PortAudio.''

3. Label sequence
In the default setting, the first/final pause is synthesized.
You can cut the pause in the label for fast response.

Regards,
Keiichiro Oura


2011/12/18 Nivedita Chennupati <niveditachennupati@xxxxxxxxx>:
> Hi,
>
> Its almost taking 6 secs. Can you please give me any suggestion to make it
> faster ?
>
> Regards,
> Nivedita
>
>
> On Wed, Dec 14, 2011 at 7:49 PM, Keiichiro Oura <uratec@xxxxxxxxxxxxxxx>
> wrote:
>>
>> Hi,
>>
>> When you start to synthesize an sentence, how much longer until the
>> sound is started?
>>
>> Regards,
>> Keiichiro Oura
>>
>>
>> 2011/12/14 Nivedita Chennupati <niveditachennupati@xxxxxxxxx>:
>> > Hi,
>> >
>> > I am currently working on a dialogue system. I have integrated HTS voice
>> > to
>> > it. But the problem is that its taking too much time to synthesis it. We
>> > have a powerful system and I think that is not the likely problem for
>> > the
>> > synthesis. Following are some queries I have regarding HTS
>> > 1. A solution I was trying to find on internet is hts server (something
>> > like
>> > festival server). Is it possible and how?
>> > 2. Which one is faster: limited domain festival TTS or HTS?
>> >
>> > Could you also suggest me some tips in using HTS for real time systems,
>> > to
>> > speed up the process.
>> >
>> > Thanks,
>> > Nivedita.
>> >
>>
>