Closed bertfrees closed 1 month ago
As part of this issue, we will also do some refactoring and eliminate some XSLT code. The voice-family property will be parsed in Java to a VoiceFamilyList
object, and this object will be passed to VoiceManager.findAvailableVoices()
. The SSML will only have to contain "voice-family" attributes, instead of a combination of "voice-gender", "voice-age", "voice-selector1" and "voice-selector2".
Possible values of the voice-family
property are:
male
, female
or neutral
), optionally the age (child
, young
or old
), and optionally the preferred variant (1
, 2
, etc.).initial
: the CSS spec says the default value is implementation dependent, and we choose to make "initial" equivalent to an empty list of component values, which means that the VoiceManager makes its selection based only on the locale preserve
: we will not support this value for nowinherit
Pipeline supports
voice-family
, but a custom version of it. Make it compatible with standard CSS.See https://www.w3.org/TR/css-speech-1/#voice-props-voice-family