eSpeak support in Orca -- what is the best way?

[Henrik Nilsen Omma <henrik ubuntu com>]

Hi all,

I've been working with Gilles Casse of Oralux on a spec for better 
multilingual speech support in Ubuntu, and as it happens, the crux comes 
down to support for eSpeak in Orca. Let me explain ...

The aim of the MultilingualSpeechSynthesis spec is to extend our current 
provision to synthesised speech in multiple languages right on the CD. 
That is not possible with Festival because the voices are too big, but 
should be possible with eSpeak. See: 
https://wiki.ubuntu.com/Accessibility/Specs/MultilingualSpeechSynthesis

We also plan to improve the speakup support on default systems and Live 
CDs using eSpeak, which fits in well with the added language support. 
However, the main focus of the Live CD is still going to be the Gnome 
GUI. So we have to support both interfaces, and we want to do it with 
the same speech synth to avoid duplication.

But we won't move the Live CD from Festival to eSpeak until we are 
confident that there is good support for eSpeak with Orca. (btw, many 
people will still prefer Festival or other synths and we should have 
good support for those and make sure installing and setting up is easy)

I know I'm probably stirring up a old debate when I ask what the best 
way to do that is. I guess there are two options:

* Write a gnome-speech driver for eSpeak -- How much work is involved 
with this? Gilles says he is willing to start on this.
* Speech Dispacher support for Orca -- I know there have been issues 
raised about this before. Some missing features are mentioned here: 
http://live.gnome.org/Orca/SpeechDispatcher

-- Using the Orca -> gnome-speech -> SpeechDispatcher -> eSpeak chain is 
not really an option for a stable release I think.

I'm not really technically qualified to have a firm opinion about which 
route is best or easier to implement. I simply note that a solution is a 
prerequisite for the multilingual Live CD and the enhanced speakup 
support. In principle I'm a fan of the speech dispatcher approach 
because I feel it open up more options for the future such as Orca 
running on KDE, but if the missing features there mean holding up a spec 
like multilingual support for a cycle or more then I'd like to consider 
alternatives.

I've made a spec describing what we need and briefly mention the two 
options.
https://wiki.ubuntu.com/Accessibility/Specs/OrcaEspeak

Another question is whether eSpeak itself is feature complete enough 
(does not  support asynchronous calls ATM AFAIK), but this is mediated 
by the ability to install Festival or something else post-install. I do 
wonder how the user community would react to a sudden switch of default 
synth though. Thoughts?

Discuss :)


Henrik