[hts-users:03125] Re: Parameters of Festival cmu_us_slt_arctic_hts voice

[Keiichiro Oura <uratec@xxxxxxxxxxxx>]

P.S.

I think that I used the 32kHz data files, the prototype of HMgeTool, etc.
Anyway, I don't remember the scripts in detail.

Regards,
Keiichiro Oura


2011/12/5 Keiichiro Oura <uratec@xxxxxxxxxxxxxxx>:
> Hi,
>
> The parameters are not publicly available.
> I don't know the performance of them for another voice.
>
> Regards,
> Keiichiro Oura
>
>
> 2011/12/4 Mark Huckvale <m.huckvale@xxxxxxxxx>:
>> Dear Keiichiro
>>
>> Thanks. Is this recipe for the analysis/training parameters publically
>> available? Is it so specific to SLT that it would not be good for another
>> voice?
>>
>> Best
>>
>> Mark
>>
>>
>> On 04/12/2011 12:21, Keiichiro Oura wrote:
>>>
>>> Hi,
>>>
>>> The SLT voice in Festival 2.1 was trained by using the standard
>>> programs and data files.
>>> But, the analysis/training parameters are very different from the
>>> standard demo scripts.
>>>
>>> Regards,
>>> Keiichiro Oura
>>>
>>>
>>> 2011/12/4 Mark Huckvale<m.huckvale@xxxxxxxxx>:
>>>>
>>>> Dear List
>>>>
>>>> I have built an HTS voice from the SLT ARCTIC audio files as per the
>>>> demonstration recipe, but the voice doesn't sound as good as the SLT
>>>> voice
>>>> that comes with Festival 2.1. It sounds more "buzzy" and has some harsh
>>>> fricatives.
>>>>
>>>> Is a different recipe required to get the superior voice, or is that
>>>> quality
>>>> not attainable using the standard programs and demo data files?
>>>>
>>>> Thanks
>>>>
>>>> Mark
>>>> --
>>>> Mark Huckvale
>>>> UCL Speech, Hearing&  Phonetic Sciences
>>>> www.phon.ucl.ac.uk/home/mark
>>>>
>>>>
>>>
>>>
>>
>> --
>> Mark Huckvale
>> UCL Speech, Hearing & Phonetic Sciences
>> www.phon.ucl.ac.uk/home/mark
>>
>>