Re: [orca-list] New Linux user, needing higher quality speech...



why can't we use decTalk for linux? if i remember right we could 8 to 10 yearrs back. DecTalk is pretty good I think.

follow me on twitter @joshknnd1982

On 5/15/2015 8:37 PM, Kyle wrote:
According to B.Henry:
# Kyle, I understand your point of view, but this kind of post is really
# once again preaching in my opinion. Those who can and wish to go FOS or
# bust are already doing so, and those who can't or won't switch away from
# Voxin will not be convinced.

I'm not preaching, I am only attempting to point out the technical
problems with using something so old and outdated that has to maintain
more and more very old compatible C libraries on its own in order to
work at all, and is unable to be rebuilt against the newest libraries
that all distros are using now. There's also the fact that a simple typo
or OCR mistake can crash the entire screen reader, and that's no
exageration. There is also a literary term, which is also the name of a
Bitcoin client written in Python that will also crash the screen reader
when speaking it is attempted. I will not write those words here, as
common though they are, I will certainly be accused of being malicious
if I write these perfectly normal misspellings here. Believe me, just
stick an h where the n should be in Wednesday and you will feel the
burn. Now go Google C a e s u r e or w e b h e s d a y. One is a
literary term that is also a Bitcoin application, and the other is a
common OCR mistake found especially in old newspaper headlines. The
first one I mentioned is a one-letter typo, where the letter typed is
very close to the correct letter, so is extremely easy to mistype. Now
tell me again that I'm preaching. I can't recommend something that is
broken, and will refuse to help someone get it going, knowing that they
will have trouble with it. Sorry Christopher, I know you'll wax
defensive over these statements, though they are indeed true. Do you
work with/for the person who continues to package this mess? If so, then
I'm sorry for you indeed, but there's still no need to be defensive of
something so defective.

<snip>

# Also voxin works on several distros, and very popular ones out of the
# box, e.g. Arch, Manjaro, Debian...

Sure it works, for now, because the packager has put in more and more
old system libraries in an attempt to keep it working. There is no
rebuilding of the code, which is either lost or obfuscated beyond
usability even by the packagers and license resellers. The old system
libraries that have to be used to make it work can actually pose
security problems as well.

<snip>

# If it is important to someone to try and get people using something
# other than eloquence, aka Voxin, aka viavoice... then I think making
# good new voices for existing synths such as espeak will be helpful. I've
# made a few, but I am not sure whether or not  they will help many people
# who have a hard time hearing existing espeak voices.

There are already many attempts at various voices for Espeak, because
making them is easy. It's now only a matter of packaging them to work
with speech-dispatcher until such time as it supports proper voice
variants, which is hopefully on the horizon to be implemented sooner
rather than later.

# The bottom line is that quality is a subjective term, and although
# certain aspects of quality can be objectively measured at the end of the
# day we are talking about personal prefference. Saying that eloquence is
# not high quality is at best a personal opinion. Again this comes from
# someone who uses ESpeak.

See above for the reasons why this synthesizer is of extremely poor
quality, and this is not just a personal opinion, it's a technical fact.
I point you once again to Googling certain words that are known among
the community of users as "crash words." Very few other voices have such
"crash words," and if they did, they would have either been fixed or
taken off the market years ago instead of being held together by the
increasing layers of compatibility glue just to keep them working just a
little bit longer. Imagine if you purchased a TV with problems like
this. Say for example you turned on your TV and found that whenever a
certain shape appeared on screen, the whole thing would crash, forcing
you to turn it off and then back on. More common would be a malformed
packet caused by degradation of the digital signal, but you get the
idea. If such a packet forced you to powercycle your TV, you would say
that it's broken or low-quality, and you would rightfully want your
money back. This is the case with Voxin, ViaVoice, Eloquence, IBMTTS,
TTSynth, and on and on and on, but it's just not getting fixed.
Therefore, the quality is very much measurable, and is very very bad,
and if I purchase any one of these, even knowing how broken it is, in
many places, I should have the right to a refund, or even a lawsuit
against the vendor for selling, and continuing to sell,  known defective
software.

# I've seen mailing lists degenerate in to spaces doninated by "my voice
# is better than yours" conversations for days on end where the clutter of
# voice prefference posts makes it annoying to look for the other content
# being posted.

This is not one of those e-mails. It is a technical summary of why
people should try to find, or maybe even make, something that works
better, even if it sounds relatively similar.

# I've sure seen more email from people who can't get festival to work
# than those who have had trouble installing Voxin, and Voxin certainly is
# easy to install on the distro the original poster is usiing.

Sure there's no trouble installing it, because someone is putting a lot
of effort into making it installable. Apparently, no one is putting so
much effort into making Festival so easy to install and use, which is
sad indeed. Voxin et al certainly aren't usable in the long term, and I
stand by this technical acessment.
Sent from my sun spot




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]