Hi Anil,
I'm thinking that if you could untie the semi-tied structure of STC and re-clustere the model set so that it could be converted to engine format? That might affect the performance of STC but would be a slight promotion?
I haven't checked the details of STC models so forgive me if I got it wrong.
cheers,
--
Na Xingyu (那兴宇)
Beijing Institute of Technology
naxy(at)bit.edu.cn
asr.naxingyu(at)gmail.com
naxingyu at {facebook, twitter, linkedin}
At 2011-09-23 13:29:05,"Keiichiro Oura" <uratec@xxxxxxxxxxxxxxx> wrote:
>Hi,
>
>hts_engine API does not support STC models.
>
>Regards,
>Keiichiro Oura
>
>
>2011/9/23 Anil John M <aniljohn80@xxxxxxxxx>:
>> Hi All,
>>
>> This is my first question to the mailing list. My apologies if this is
>> too naive.
>>
>> I have build an HTS voice for my data using the demo scripts of SLT.
>> When I observed the synthesized files of alice sentences in the gen/
>> folder, I found STC based samples are sounding smooth than hts_engine
>> samples.
>>
>> The models that are generated in voice/ directory can be used with
>> hts_engine_api with respective switches for real-time synthesis. But I
>> do not know where the STC models will be generated and how to use
>> those models for real-time synthesis. I will be helpful if any one can
>> provide this insight with a simple example.
>>
>> Thank you,
>> - Anil
>>
>>
>