[LAU] Synthesized voices [was :Re: [LAD] Kontakt sampler format (and others like EXS24)]

[Patrick Shirkey]

On Mon, September 3, 2012 12:44 pm, SxDx wrote:
>> From: "Lorenzo Sutton" <lorenzofsutton at gmail.com>
>
>> >>>> From: "Ralf Mardorf" <ralf.mardorf at alice-dsl.net>
>> >>>> The East West Choirs sing your lyrics. It's not another Ahhhh or
>> >>>> Ohhhh
>> >>>> sample ;).
>> >>>>
>> >>>> http://www.youtube.com/watch?v=aI5Gg2-mhmU
>
>> I think you only need the resources/time/skill to sample all the
>> phonemes for various choir combinations, add glue and logic, the
>> interface, some usual ADSR stuff and filtering, reverb here and there, a
>> pinch of randomness, and your done :)
>>
>> Another way could be to have 4 singers sing the lines various times
>> recording them multi-trak, each take with different miking positions and
>> EQing (or keeping the mike still in the room and having them move around
>> each time). If the singers were good enough they'd be able to change
>> their voice quality slightly at each take to mimic multiple people. Of
>> course you need 4 singers if you wanted a full SATB choir, probably just
>> a good baritone and dark soprano would suffice to have a mixed choir
>> effect for a monophonic or 2-voice line.
>
> Maybe that's how they do it in the video. I gave a try to mbrola and the
> result is much more realistic than espeak. Mbrola uses recorded
> phonemes (if I understand correctly).

Yes, and that is the same technique the vocaloid uses too.

There has been a lot of funding and research undertaken by Inria and at
one point they even got a song onto the charts ;-)

Festival also has a singing mode and it can use espeak or mbrola.

I'm sure in a couple of years there will be a large database of singers
for the lauloid project. If we took it a step further lauloid could be
integrated with linuxsampler too.




--
Patrick Shirkey
Boost Hardware Ltd