gender voice variants

[Phillips, Addison]

I didn't say that there was a transformation involved. For TTS/STT applications there is, but not for other cases, such as static voice recordings. My suggestion was meant to point out that the information you're suggesting (for use in language ranges and also language tags) is language-related (like what extensions cover), but not necessarily descriptive of the *language* itself. And the tags you're suggesting are, depending on the application, probably incomplete. Consider:

"en-US-x-male-en-GB-adult"

There are a variety of things that go into selecting voice/audio assets or for generating of audio from text at runtime. Language tags might not be the best mechanism for doing voice personalization--or a myriad of other personalizations.

Addison

> -----Original Message-----
> From: Peter Constable [mailto:petercon at microsoft.com]
> Sent: Tuesday, December 18, 2012 9:46 AM
> To: Phillips, Addison; Patrick; ietf-languages at alvestrand.no
> Subject: RE: gender voice variants
> 
> There's no transformation involved in an audio recording, and may not be any
> kind of transformation involved in a variety of speech assets.
> 
> Peter
> 
> -----Original Message-----
> From: Phillips, Addison [mailto:addison at lab126.com]
> Sent: Tuesday, December 18, 2012 8:18 AM
> To: Patrick; ietf-languages at alvestrand.no; Peter Constable
> Subject: RE: gender voice variants
> 
> Patrick wrote:
> >
> > What about languages distinguishing between female and male speech
> > such as, apparently, Yanyuwa in Australia? Or Japanese onnarashii
> > (女らしい),
> > aka speech for ladies?
> 
> Those would be *orthographic* variations with-in *specific* languages. These
> might (or might not) rise to the level of variation that requires separate
> identification via a subtag. But that's different from what Peter was asking
> about, which was identification of a variation in the audio reproduction of
> *any* language. I can even think of some good use cases where such subtags
> would be useful--for example, identifying different pre-recorded audio
> resources that vary by gender.
> 
> However, I still don't think that this would be a good use for the language
> subtag registry. Perhaps something like the transformation extension (which
> does describe the changes to language material as the result of some external
> process)?
> 
> Addison
> 
> Addison Phillips
> Globalization Architect (Lab126)
> Chair (W3C I18N WG)
> 
> Internationalization is not a feature.
> It is an architecture.
>