W3C home > Mailing lists > Public > www-style@w3.org > January 2011

Re: [css3-speech] pitch alteration expressed in semitones and Hertz

From: Daniel Weck <daniel.weck@gmail.com>
Date: Wed, 19 Jan 2011 23:05:52 +0000
Message-Id: <58B7B393-AC28-4489-B0C9-96737CE179EB@gmail.com>
To: www-style@w3.org
Thanks for your thoughts Fantasai.
SSML disambiguates absolute pitch values (in Hertz) and relative  
increments/decrements (in Hertz or Semitones) by forcing the use of  
the "+"/"-" prefix before the number. In CSS-Speech, what do you think  
about forcing the use of the "Hz" suffix not only in relative values,  
but also in absolute ones? This would remove the ambiguity you are  
referring to. (percentage and named values obviously don't need such  
distinction)
Newly-produced content based on this updated specification would break  
existing implementations, but so would values written using the new  
additive/substrative syntax...so I don't see this as a huge problem  
given the 'draft' nature of the specification. Obviously we would need  
a consensus on this kind of detail before moving to Last Call Working  
Draft.
Regards, Dan

On 19 Jan 2011, at 22:33, fantasai wrote:

> On 01/19/2011 01:19 PM, Daniel Weck wrote:
>> Just in case this isn't obvious to those who are not familiar with  
>> the previous CSS3-Speech draft [1]:
>>
>> The 'voice-pitch' and 'voice-pitch-range' properties were already  
>> featured, but only with support for fixed Hertz values,
>> predefined named values, and percentage values (allowing relative  
>> changes in a multiplicative fashion). There was a pending
>> issue regarding a discrepancy with SSML 1.0, which enables relative  
>> changes in an additive fashion (increments and decrements
>> of non-percentage / fixed values) using both Hertz units and  
>> semitones.
>>
>> I addressed this issue by adding an extra "value type" for the  
>> aforementioned properties, so the current specification is
>> backward-compatible with the previous draft (none of the previous  
>> "value types" are affected by the change). Existing
>> implementations only need to update the parsing algorithm, and to  
>> add the conversion from semitone offsets to fixed Hertz
>> frequencies (which is based on trivial arithmetics).
>
> The value type you are adding would be parsed as a dimension,  
> similar to lengths.
> This means the plus sign is in fact optional.
>
> It might be worth investigating whether there are any  
> implementations of the
> CSS3 voice-pitch and voice-pitch-range properties. If there aren't,  
> you will
> have more freedom to change the syntax. Having numbers and Hz both  
> refer to
> hertz, but mean different things, for example, is imho somewhat  
> confusing.
> An alternative might be to add a keyword
>  voice-pitch: <semitones> | [ <hertz> && absolute? ]
> or somesuch.
>
> Wrt implementations, the only two CSS Speech implementations I know  
> of are
> Opera's <http://www.opera.com/docs/specs/presto23/css/properties/> and
> Emacspeak <http://emacspeak.sourceforge.net/>
>
> Opera I believe follows the CSS3 drafts, but only has partial support;
> Emacspeak claims full support for the CSS2 variant.
>
> ~fantasai
Received on Wednesday, 19 January 2011 23:06:32 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 17:20:36 GMT