SAPI in Ubuntu using Gnone Speech Dispatcher

[AICollector]

I would very much like to get my 16kh SAPI voice working in Ubuntu!

Just an idea; some tools on Windows let you use the SAPI system to read out a txt file and turn it into a .wav or .mp3 (for later listening) I used to do this to turn a few good books from Project Guternburg into audiobooks.

Any chance I might see similar functionality with this, I wonder?


It's probably also worth noting most programs like this could also allow the user to make the computer pronounce words diffrently.


When your a bit further down the line (IE: you've got an interface an idiot like me can use) I will be more then happy to help!

[notlistening]

Yeah the basics have already been done and the system will produce wav files on demand, the code is just not released yet. The prenounciation is an issue i have looked at a little but feel that it is doable as well.

Tom

[plantatree]

I would very much like to get my 16kh SAPI voice working in Ubuntu!

Just an idea; some tools on Windows let you use the SAPI system to read out a txt file and turn it into a .wav or .mp3 (for later listening) I used to do this to turn a few good books from Project Guternburg into audiobooks.

Any chance I might see similar functionality with this, I wonder?


It's probably also worth noting most programs like this could also allow the user to make the computer pronounce words diffrently.


When your a bit further down the line (IE: you've got an interface an idiot like me can use) I will be more then happy to help!

In fact there is a free editor Saypad that can read to wave and mp3 as well can split up files given a chapter break. I use it through wine to prepare books for my wife. It's a little bit of a hassle to get used to it though. As well for every file I have to restart the editor.

As well it has another drawback. Only the standard 3 sapi voices are selectable through it. But I wanted to use a separate voice because of my language. So eventually hacked the voices' names in the windows/wine registry and now MS Marry is the voice I want to read with

Experimenting with open-sapi now. Unfortunately I'm facing an issue with speech-dispatcher not processing cyrillic. http://lists.freebsoft.org/pipermail...q2/001695.html

[notlistening]

Latest developments, now open SAPI seems to work in other languages. It is hard coded at the moment so there is no flexibility on the encoding used but the proof of concept is there again and keeping a few users happy.

Tom

[notlistening]

Right this now requires testing from people. The default wine version in Ubuntu even 9.04 is not stable enough when it comes to running threads so you need to get the latest version of wine for a reliable system. Currently in the testing phase and require people who can use a terminal to get stuck in.

NL

[notlistening]

I now have a very stable version of the software running without crashes. I do not use this as my main way to access the computer as i have enough sight to use a magnifier.

It would be great again if you guys could test the system. It is important to update to the latest binary version of wine from the wineHQ website. This improves stability. Otherwise i am using the current version of speech-dispatcher and Orca in 9.04 with small modifications to the speech dispatcher setup. This can all be downloaded from the project website and there is a wiki giving installation and usage instructions.

This software is still in the development stage but coming along slowly. Please try it out and see.

Tom

[notlistening]

What's new:
Vista TTS engine installed and working
Threaded model working smoothly
Gnome Lockups fixed
Delays and freezes fixed


The release of Karmic Koala has created an opportunity for open-sapi. The latest release of Ubuntu has seen a few usability problems introduced for text to speech users. Reliability, latency issues with the audio subsystem and generally the move toward pulseaudio that ubuntu seem to be pushing for seem to be making the system unstable and unusable.

A mist these changes it has been just in time for open-sapi. With the latest development version as yet unreleased ( The setup is a real pain by the way at the time of writing this) open-sapi in its second generation is out performing espeak on karmic.

The clarity of speech is much clearer even with the basic Microsoft (free engines) it is significantly more responsive. Delays and blocking behaviours previously seen are now a thing of the past.

This has been achieved using a threaded approach but has come at the cost of having to use a more up to date version of wine. As wineHQ have not released a wine version for Karmic it meant building from source or using the Jaunty repository (untested).

This is further complicating the open-sapi install. The plan is to create my own PPA with the a stable wine version for use with open-spai until Ubuntu decide to update to a later wine version.

I have to iron out a few bugs with how the system loads and runs, package the open-sapi files into a .deb file for the first release, find a way to install SAPI and submit the first round of changes to speech-dispatcher.

Tom

[notlistening]

Latest News,

So where we are now. I have a reasonably stable open-sapi that runs with Speech Dispatcher and Orce through wine. It is fast responsive and sounds nice. Changing the voice in open-sapi is a pain at the moment and none of the Orca setting pitch, volume and rate actually take affect . If people would like a sneaky peak I can get the pitch volume and rate all working and release a pre-Alpha deb. Let me know if your interested.

Change Log:

• Fixed command line bugs with ?
• Enable file/wav format changes
• Enabled wav file output
• Streamlined Code
• Moved to a multi threaded model
• Reworked the debugger to use threads
• Reworked speech output to use threads
• Rework audio output to use Memory Streams

A recent move to Lucid has meant that open-sapi had to evolve in a new direction.

The update of wine in Lucid was a very much needed component for open-sapi to even run under Ubuntu. This now allows for the multi threaded server to run smoothly with significant performance improvements with open-sapi responsiveness and application start up times.

However with every update comes another problem. We are supposed to call it progress.

The wine ESD sound driver is soo out of date that pulse audio and wine are no longer best friends and the audio is sometimes distorted sometimes not, with lots of pops and fizzes on the output. Also from what I read this is different depending on your hardware. So using the wine audio subsystem is not really viable anymore for open-sapi.

So I have begun development to get RAW audio data from sapi and pass that to the client who can do what they want with it. The proof of concept is done and I have RAW audio streams that I can process, redirect anyway I want. This was a big project milestone as it was a stopper on the development of the Speech Dispatcher module. Also it allows me to follow on to actually creating a speech server which can output multiple streams to multiple clients at the same time.

NL