SAPI in Ubuntu using Gnone Speech Dispatcher

[notlistening]

Hi Everyone,


I have been considering a project for sometime now and have finallly freed enough time to begin working on it. It is really a quick hack to get the MS SAPI speech engines to work under Linux not just Ubuntu.

I have SAPI running under WINE *Note that Hardy breaks the install process of the MS SAPI 5.1 SDK. I have hacked that around a bit and got it working again but by default it does not work. I have submitted a bug report to the WINE team.

I also have VWKate SAPI 5 from NeoSpeech installed. This has been tested using the TextAloud package and also the MS examples under WINE.

From there I am in the process of writing an executable in windows that implements the basics of the SAPI engine and Gnone Speech Dispatcher interfaces.

To increase the speed of development (in theory ) I have used tcl to script the SAPI to Speech Dispatcher middle ware. This is using a combination of TCOM for windows com objects into the SAPI engine. Speech Dispatcher uses standard in and out for communication.

This will create a Server Client Model onto the SAPI engine that can allow for communications from Linux based programs and software to call the SAPI engine to speak.

So I think it is possible to get this functioning and I am interested to know from people here:

Firstly if people can get over the fact that i am using MS on Linux.

Secondly if there is anyone with any suggestions, that wants to help out, that have a particular interested to see this working. Please drop ne a message or post on this thread.

Tom

[doviende]

Yes! I will worship the ground you walk on if i can use SAPI5 voices in linux. I haven't yet figured out how to do this...i tried installing textaloud in wine but it can't see the sapi5 voices i have, so i'm stuck.

i was considering running windows in vmware to try to make this work because i desperately need it right now.

[notlistening]

Proof of concept it works! Hopefully we will see SAPI on Ubuntu soon.

I now have an executable that speaks using SAPI using wine and the basic Microsoft default speech engine. I have also installed an additional speech engine from NeoSpeech VWKate and once selected it will use any installed speech engine.

Main Tasks:
1. Speech Dispatcher SAPI driver
2. SAPI Installation & Config
3. Middleware
4. Package

ToDo Breakdown:
1. Create stdin file event handeler - Not Started
2. Create dispatch code for stdout - Not Started
3. Rate, volume, voice, sound device - In Progress
4. Event reporting - Researchig
5. Read SPI specification - Researching
6. Create aware network enabled option - In Progress
7. Intergrate tcom into standalone executable - Done
7a. Create install for C++ library to /system32 - Not Started
8. Install SAPI on Ubuntu - Broken but Jimmied - In Progress
8a. Create standalone installation package - Not Started
8b. Identify and generate filelist - In Progress
8c. Identify SAPI registry entries & Reproduce - In Progress
9. Test SAPI 4/5 Install Packages - 4 Tested
- Default MS Engine SDK5.1 - Working
- VWKate Neospeech - Working

Any help still welcomed, especially testers

Tom

[Ohmu]

WOW.
W O W
Tom, You're a legend! This is incredible! Amazing!

I'm just looking at doing exactly the same thing, Was getting desperate. Then I see your post!

I'm pushing to get a wholly FOSS speech engine+GUI.

I blogged the start:

http://womblezone.blogspot.com/

<exerpt>

The plan

1. Use WINE to get Vista's Speech Engine operating in Linux
2. Create a GUI that'll interface with this engine.
The GUI will sporadically (unless the user disables the feature) send phrase-data to a central database (say VoxForge - I have contacted the maintainer and he is friendly)
3. Once we have enough data, throw out the WINE-wrapped Vista Engine, and replace it with our own FOSS engine.

Please have a look and tell me what you think!

hmm SAPI 5.1 doesn't have a UK english voicemodel. This is why I was trying to port the Vista one, but it's possibly horribly embedded. Anyway, I'm well up for learning American, in order to write my emails.

I hope we can meet up on the internet (I'm in India now) and chat! How about irc.freenode.net#cmusphinx?

Let's do it! Let's get a SAPI engine in Linux, and seed a FOSS GUI that kicks Vista Speech Assistant's bony butt!

Sam

[Ohmu]

Quick update,

I'm now running Vista + Visual Studio in VirtualBox

I'm using the SimpleDictation sample from the Windows SDK

...compiling a release build and trying to port this to Wine. It's the simplest possible way to test that SAPI 5.3 has transferred over ok.

I think there's a good chance of this working... I'll post another update in a few days.

Sam

[notlistening]

Hi Sam,


The initial work that i have done was proof of concept and it works under SAPI 5.1 but with quite a large amount of complex sticky tape work. The current version of Wine does not play niceley with the SAPI installer.

Microsoft have been a little bit sneaky. They have included the SAPI 5.3 SDK files within the Vista SDK package. To work with the Vista SDK you will need to identify and extract the required files and make your own install package for Wine. Similar to the process that i will be using. I see the work that you are doing as the next natural step and would welcome you to make a start.

I am using some less popular technologies and would not expect you to use them but they do provide the basis for a working model and test environment.

I have not looked at the legality of using either SAPI 5.1 or 5.3 might cause when used in this way.

Oh well if it is popular then it will catch on.

Hope to catch up soon.

Tom

[Ohmu]

Check this link!

http://www.geekpedia.com/Thread16235...cognition.html

This guy has ported SR from Vista to WinServer08. He details the files and registry settings that need to be copied over.

I'll copy the relevant part of the page here, to keep things together.

Okay. So I DID get it to work but I can't tell you the exact steps but I can
tell you what I did and what I think matters. Hopefully someone else can
narrow this list down to what counts. Or even better, someone from MS support
can actually reply to one of the 5 messages posted. What you need: Machine A
with server 2008 and Machine B with MS Vista (FYI - my vista is NOT on SP1
yet so that shouldn't make a difference)
NOTE - ALL copies were done w/out writing over current files, I left
existing items in place.

1. Copy from vista(a must) C:\windows\system32\speech
2. Copy from vista(a must) C:\windows\speech
3. Copy from vista (prob not needed) C:\windows\assembly
4. Copy from vista (prob not needed)
C:\Windows\Microsoft.NET\Framework\v2.0.50727

Registry Changes (do at own risk but this is the most crucial part) Now, I
am not sure how classids are created but i THINK (please correct if wrong)
they are unique to each machine which i think caused issues for me but if I
am wrong, and I hope I am, someone can actually provide a list of which ones
matter.
1. Export from vista(a must) HKLM\Software\Microsoft\Speech
2. Export from Vista (optional?) HKCU\Software\Miscrofot\Speech

Now the hard part. There are references of GUIDS that matter (its the
objects that look like this: {DAC9F469-0C67-4643-9258-87EC128C5941}
Find everyone one of those under the registry objects you just exported in
Vista and export to server 2008. Really, the ones I think matter are the HKLM
and HKCR. If you did the current user then you should continue and do the
HKCU as well. This is the most tedious part, but i think its a must.

Do not overloook this-> HKLM\Software\Classes\CLSID just because you did HKCR

And lastly, (this is what I think finally got it workin). On vista look for
every registry reference to the following files and bring those objects over
(mainly its the classids again inside of HKCR and
HKLM\Software\Classes\CLSID).

C:\Windows\System32\Speech\Engines\SR\spsreng.dll
C:\Windows\System32\Speech\Engines\SR\spsrx.dll
C:\Windows\System32\Speech\Engines\SR\srloc.dll

Other actions I took or variables that I don't think matter, but I wanted to
share. My vista OS is ultimate (but ANY vista should work). I had installed
speech recog server 2007 (which you can download from MS) on server 08 and
vista. I installed windows SDK for server 2008 on both machines and compiled
and registered the sample recognition engine (which doesn't do anything).

I think that covers it. It was a horible process that most likely people
will not want to attempt, but i promise you it is possible. I think its all
about know which reg keys to export and the basic files. I hope MS releases
something to make it easier because it should be a very simple update for
them. If my server crashes i would be very bummed. (now to take an image of
my server)

[drbongo]

Greetings 'notlistening' drbongo here!

I was going to recommend that the people interested in getting sapi working on ubuntu with wine get in touch with you, but it seems that you are already contributing to the thread. I have just signed up for the accessibilty team and found your tasteful 'brokeback mountain' style picture which I presume was taken during the breast cancer awareness week!

I am currently working on remastering Ubuntu with the accessibility options enabled. Initially three versions one with speech, one with magnification and one with both. Then I will need to test as many applications as possible to see what can be done and what can't.

Keep in touch!

drbongo

[notlistening]

Hi Dr.

I started the post a bit ago and have got very little interest. Lots of people want it but there are not to many people who are interested in working on it. I am setting up my website http://www.vipsight.info to host some information, guides and downloads. If you have a spare 65 MB and use speech dispatcher then I can do you a SAPI install. Soon I will write a few useful tool as examples but for now time is short.

Tom

[drbongo]

Hi Tom,
Keep me in the loop, as you know my 'Linux Development' time is somewhat limited during working hours as my employers want software they can sell, but I am happy to try out anything you develop or assist you if I can in my own time!

I am planning to focus on making accessible remixes of Ubuntu in the short term, In the longer term I am hoping to make some accessible apps using the Gambas IDE which will be installable on any of the major Linux distros such as Debian, Ubuntu, Fedora and Suse etc.

I will post any developments or queries to this forum which might be the best way to keep in touch.I will start a new thread when I have anything worth posting!

drbongo