VmWare & Festival, VM Benefits, CLI and Gnome Annoyances

Tue Sep 26 23:37:29 UTC 2006

Hi list,
My first post here. This is pretty long so feel free to snip and mod the 
subject line as needed.

VmWare & Festival:

I'm running Dapper under VmWare Server 1.01 (Win XP) and performed an 
alternate Ubuntu install. Audio works fine in Gnome but I get no speech in 
Gnopernicus. Of the two servers available in test-speech the second gives me 
an error and the first place corrupted audio sounding like bursts of white 
noise and other artifacts. Why is there no speech in Gnopernicus even though 
Gnome sound does work? What audio attributes does Festival have? I'm 
suspecting the audio artifacts might be either some issue within VmWare 
itself or some non standard audio format e.g. odd sampling rate used by 
Festival. I'd appreciate any feedback you can give me about troubleshooting 
Gnopernicus speech issues and running Festival in VmWare.

Here are some extracts from test-speech:

Attempting to activate 
OAFIID:GNOME_Speech_SynthesisDriver_Festival:proto0.3.
Driver name: Festival GNOME Speech Driver
Driver version: 0.3
Synthesizer name: Festival Speech Synthesis System
Synthesizer Version: 1.4.3
<snip> This one plays corrupted audio.

Attempting to activate 
OAFIID:GNOME_Speech_SynthesisDriver_Speech_Dispatcher:pro
to0.3.
Server could not be initialized.

A reply I got in alt.comp.blind-users seems to indicate that the audio issue 
might be something introduced in VmWare Workstation 5 and the Server product 
line. The post is here (via Google):

http://tinyurl.com/rcd5z

I've also reported the issue, together with a sample wave file, on the 
VmWare Web fora but got no reply:

http://tinyurl.com/hbk9m

I went with the defaults during the OS install itself and the only 
customization I've done has been to enable assistive technology support and 
Gnopernicus in Gnome. I also ran all the 30 updates it recommended including 
apparently a new Linux kernel. I haven't found any useful audio options in 
VmWare or the GNoem audio applet itself. The guest machine's sound card is 
always an SB compatible PCI card which shows up as Ensoniq PCI audio within 
Linux.

If Festival fails to work, VmWare offers a named pipe back to the Windows 
world from its serial port. Are there any simple speech APIs that 
Gnopernicus supports? If there are I suppose I could write a quick and dirty 
Perl script that sits at the other end of that named pipe waiting for input 
and passing it after some wrapping to a SAPI 5 OLE object in Windows. That 
means WIndows speaks the prompts in Gnome but I don't care which OS does it 
as long as it works, <smile>.

VM Benefits:

My method of installing Linux was pretty involved but has been working quite 
well actually. That is I ran the alternate Dapper ISo selecting the 
blindness boot option and giving the parameter console=ttyS0. That guest 
machine's serial port is a virtual Windows serial port whose connected to 
the TeraTerm terminal emulator. I can then use my screen reader and 
magnifier, Supernova, to read all terminal output.

Although it required some configuration, I kind of like my current assistive 
technology setup. It means I can use a screen reader straight from the boot 
up, regardless of which distro I'm trying to run or whether sound works in 
the first place. As VmWare is a virtual machine I can reinstall everything 
with ease, take snapshots and even use full-screen magnification and color 
changing (invert RGB) features provided by Supernova. How's Gnome 
magnification today? The last time I tried a couple of years back it 
couldn't do full-screen magnification or magnify the area covered by the 
magnifier window.

Anyway, within my VmWare setup Braille works directly, too, and I find I can 
listen to formant based speech synths like Orpheus or ViaVoice twice as fast 
as I can Festival which is a huge productivity boost. Although I do all of 
my computing in English, I need FInnish speech support on a daily basis for 
e-mail, e-books and such. I've never seen the Suopuhe addition to Festival 
being integrated in any Linux distro. More info at:

http://www.ling.helsinki.fi/suopuhe/english.shtml

The reason why I'd like to get Festival working would be to have some spoken 
output in Gnome, too, in addition to the magnification. I'm still primarily 
a speech user. I've got a page about my sight here:

http://www.student.oulu.fi/~vtatila/sight.html

And another comparing synths like Festival and Orpheus:

http://www.student.oulu.fi/~vtatila/reviews_of_speech_synths.html

I'm considering taking part in a Unix programming course which is about GUIs 
so I'd need to get this stuff solved before I can attend.  I've tried half a 
dozen distroes over the years and in none of them the assistive technology 
has worked straight away. Most of my Linux experience has been trying to set 
up them in one distro or another, rather than actually using and enjoying 
the OS. In contrast, although it might be an unfair comparison, VoiceOver 
works straight away which is a great feeling if you are a newbie in some OS 
and pondering whether to give it a test drive.

CLI and Gnome Annoyances:

Some of this last section is slightly OT. It also serves as a member intro, 
though, so it's stuff I would have said sooner or later.
Ever since having seen my first GUI, I've been a GUI guy despite being 
legally blind. I value the command line and scripting occasionally but enjoy 
immensly the hotkey driven, mnemonic and simple GUI when wearing my user 
hat, period. It's gotten up to the  point whre I'm afraid to delete stuff 
without an undo mechanism like the Trash and gripe about commands whose 
names or long switches would be considered non-intuitive if they were names 
in, say, Java code.

I guess you get the picture as to my computing habits and attitude. I should 
add I consider myself a power user in the MS OSes (DOS since DrDos 4 and 
Winblows since 95) and like tweaking only if it has got some true 
significance e.g. shortening the tab order or less latency in soft synths. 
The few text-oriented Unixisms I like are Perl, I'm willing to take 
complexity in programmer mode, and  regular expressions for navigating text 
files.

Another point is that the command line isn't always that speech friendly, 
either. AT the risk of a tangent, ls -l output is extremely speech hostile 
as it doesn't place the columns in the order of signifnicance, for example. 
Similarly bad are man pages, as they wrap in mid-word and trip up the 
pronounciation in line-based speech synths.  I do realize these tools are 
not set up for speech users by default but it would be nice if more 
configuration options or accessibility scripts would be supplied to avoid 
having to reinvent the wheel.   As far as shells go, I think Fish might be 
the one for me. Is it any good with speech?  Here's some info:

http://arstechnica.com/articles/columns/linux/linux-20051218.ars/2

Lastly, there are some accessibility things I'd really like to change in 
Gnome. To my knowledge none of the high contrast themes offers a good 
contrast between the dialog, text field and button backgrounds. This is 
extremely important to me in getting the big picture without magnification. 
An example of the core rules I use when doing GUi themes for myself is my 
Accessible Winamp Skin page at:

http://www.student.oulu.fi/~vtatila/winamp_access.html

For another working  example, here's a CSS file I use to override all colors 
on the Web on WIntel platforms:

http://www.student.oulu.fi/~vtatila/ie_css.html

Again my Linux theming experiences are a couple of years old. Is there 
already a graphical theme editor for Gnome which has a real time preview? I 
find that I don't often get the look right the first time so a GUI would 
save lots of time and nerves compared to handcoding in XML or some Gnome 
specific mini language. The last time I tried setting up this in Debian I 
found no proper docs about the format, and decided to wait until they've got 
a GUI for managing the GUI, like KDE does.

Hope I've been outlining my Festival problem clearly
and that my depressing experiences and attitudes about LInux haven't boared 
you to death.

PS: I wouldn't want to start yet another holy war about GUis and CLIs or any 
other such topic. I just decided to mention right away I don't feel 
comfortable with CLIs and am aiming for a GUI experience if at all possible. 
OS X is high on my gotta get OS list.

-- 
With kind regards Veli-Pekka Tätilä (vtatila at mail.student.oulu.fi)
Accessibility, game music, synthesizers and programming:
http://www.student.oulu.fi/~vtatila/ 