[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[gnuspeech-contact] Re: GnuSpeech Text-To-Speech Server
From: |
Dalmazio Brisinda |
Subject: |
[gnuspeech-contact] Re: GnuSpeech Text-To-Speech Server |
Date: |
Tue, 6 Jan 2009 11:48:26 -0600 |
Hi David,
Just an update. I've temporarily resolved the punctuation issues. The
issue was that the TTSParser class in the GnuSpeech framework was not
quite finished as there are still a few clearly marked TODO's in the
source documentation. Until these can be looked at more closely, I've
just added a quick fix to provide more flexible handling of
punctuation so that all punctuation-ladden text will now render
instead of resulting in a "bzzzzt" synthesis error embedded within the
speech from pronunciation lookup failure. I've clearly marked those
minor changes to indicate it is only a temporary fix.
Clearly, someone has a sense of humor... ;)
Best,
dalmazio
On 6-Jan-09, at 2:10 AM, Dalmazio Brisinda wrote:
Hi David,
Some good news. We've got a basic text-to-speech server operational.
I've extracted the required bits from Monet and placed them in their
own server process using the Distributed Objects architecture, and
created this server as an OS X LaunchDaemon (the recommended way).
So there is now a simple Distributed Objects interface and an
example client application that shows how to invoke the speech
synthesis functionality programmatically. Will fill out the
functionality on the server shortly though, currently only the
"speak text" functionality is operational.
Also, I've created a Service for the speech synthesis functionality,
so it can be accessed from any application on OS X. Just select the
text in the application in question, go into the application's main
menu and click Services->GnuSpeech->Speak Text. (Incidentally,
Apple's own speech service is available at Services->Speech->Start
Speaking Text).
Now, I haven't done any work on the actual parser. I noticed that
whenever there is punctuation in the string adjacent to the word, it
treats it as a single word, and there is no entry for such a word
+punctuation. It looks like this portion of the parser/filter was
temporarily turned-off due to some complexity/incompatibility in
the way codes were inserted into strings. I'll see if I can have a
closer look at this shortly.
I've committed everything so far to the repository including
installation instructions (README.txt) for the time being, as
installation of the GNU text-to-speech server and service are not
especially transparent. That is, until an automated installation
process can be fashioned.
Best,
Dalmazio