[DECtalk] List UP?
Alex H.
linuxx64.bashsh at gmail.com
Mon Aug 29 22:17:26 EDT 2011
Hi,
No, but they just have a tendency to mess with the punctuation and
sort of commandeer it to their will, not taking into account that
somebody might have a clever synth like DECtalk. Anyways, I hear ya on
the abreviation issue. My iPod says 25 millibars instead of 25 MB for
filesizes. Or how about Windows 98 Southeast? Hahah!
Alex
On 8/29/11, Brent Harding <bharding at doorpi.net> wrote:
> I know, the automatic expansion of abbreviations drives a person up the wall
> trying to do critical things like medical transcription. I was listening to
> Jonathan Mosen do a demo of some self-voicing daisy editing app that FS
> bundles with the Plextalk, and kid me not, files sizes are measured in
> kerobytes. Is that what Microsoft does when you put kb next to a number?
>
> ----- Original Message -----
> From: "Alex H." <linuxx64.bashsh at gmail.com>
> To: "DECtalk Discussions" <dectalk at bluegrasspals.com>
> Sent: Monday, August 29, 2011 7:45 PM
> Subject: Re: [DECtalk] List UP?
>
>
>> Hi,
>>
>> Just to clarify my earlier post, life would be great if Microsoft
>> didn't feel the need to muck about with how SAPI5 handles punctuation
>> and I'd love a singing sapi. The reality and how MS has done this
>> really makes it a royal pain, at least as far as I can tell with
>> little to no programming knowledge and just reading the documentation
>> on how this stuff works. The great thing if we were to have a modern
>> speak window for singing and skits and the SAPI for scren reading and
>> other text to speech programs is that it'll still be easy to export
>> songs and skits as wave files with minimal effort and ensure that this
>> program will run in Windows operating systems in the future.
>>
>> Thanks Ed for all the tweaking and coding that has to go into a
>> project like this. It's much appreciated. Things are looking up.
>>
>> Aelx
>>
>> On 8/29/11, Jake McMahan <mcmahan.jake at gmail.com> wrote:
>>> To me, I think the whole dad gum thing is extremely impressive and it's
>>> coming along. I am looking forward to getting my hands on this thing
>>> when
>>> it is available, and using it in my DSpeech scripts and much more.
>>>
>>>
>>> From: ebruckert Bruckert
>>> Sent: Monday, August 29, 2011 6:09 PM
>>> To: DECtalk Discussions
>>> Subject: Re: [DECtalk] List UP?
>>>
>>>
>>> Okay sounds like you've done a thorough job. But I have two programs say
>>> the
>>> command line and batch mode program and speak which are both freely
>>> distributable in including source code and could certainly be included in
>>> any distribution.
>>>
>>>
>>> On Mon, Aug 29, 2011 at 3:43 PM, Alex H. <linuxx64.bashsh at gmail.com>
>>> wrote:
>>>
>>> Hi Ed,
>>>
>>> I've done some digging about how to make the SAPI5 version of DT sing.
>>> It won't be easy to do this I'm afraid. Too many variables involved,
>>> and no program except ttsapp.exe will pipe the exact string you send
>>> to it to the SAPI5 engine. That's only part of the issue though.
>>> Firstly, SAPI5, as I thought, uses some low level punctuation
>>> filtering that comes even lower in the stack than things like text to
>>> MP3 programs, like Verbose or TextAloud, etc. Basically, the text to
>>> speech program (like textAloud) gets the text and adds its own
>>> punctuation rules to the text and sends this to SAPI, and SAPI does
>>> its own filtering/changing of the text string, so the original text
>>> gets lost early on in the chain. There's not any real easy way around
>>> this I'm afraid, after a good few hours reading MSDN articles and what
>>> not. Even when talking about changing text to be pronounced better (we
>>> could use this to make DT sing), SAPI unfortunately makes its own
>>> rules of phonemic definitions when you specify the proper tags [3].
>>> This basically means that it tries a one-size-fits-all approach and
>>> some synths that are SAPI compliant still won't speak the text as the
>>> SAPI phonemic string intends. In short, this would be a nightmare to
>>> implement using SAPI5. At least the way I see it.
>>>
>>> I'm thinking for the ease of use factor for singers, maybe the best
>>> idea is a SAPI dectalk installable synth that works wiht screen
>>> readers, and then a speak window bundled wiht it for singing. That
>>> way, nobody is messing with sapi protocols.
>>>
>>> Sources:
>>>
>>> [1]
>>> http://msdn.microsoft.com/en-us/library/ms719574%28v=vs.85%29.aspx
>>>
>>> [2]
>>> http://msdn.microsoft.com/en-us/library/ms717252%28v=vs.85%29.aspx
>>>
>>> [3]
>>>
>>> http://elearningtech.blogspot.com/2010/10/using-punctuation-and-mark-up-language.html
>>> [4]
>>> http://msdn.microsoft.com/en-us/library/ms723638%28v=vs.85%29.aspx
>>>
>>> Sorry for the hugely complex post, but I felt it was needed to explain
>>> further. As always, great work and i look forward to more testing.
>>> Alex
>>>
>>>
>>> On 8/29/11, jake mcmahan <mcmahan.jake at gmail.com> wrote:
>>> > On 8/29/2011 3:07 PM, ebruckert Bruckert wrote:
>>> >> These are all generated by just plopping the text into speak, since
>>> >> the scaling and interaction of these things has always varied
>>> slightly
>>> >> the the goal now is to try to adjust things the detailed stuff like
>>> >> little overloads for the last step.
>>> >> I'm not sure if I reset Paul or left them at the custom voice when I
>>> >> did the weather, so I'll resend the weather being sure it's regular
>>> >> Paul, but in fact if everybody likes the style we can modify those
>>> >> base parameters, as I already know that the pitch scaling is
>>> slightly
>>> >> different so will probably have to change the pitch things but
>>> that's
>>> >> the purpose of tweaking. On some of the voices that I fooled around
>>> >> was so much I really don't have a feel for what's better or worse.
>>> >>
>>> >> On Mon, Aug 29, 2011 at 2:52 PM, Alex H. <linuxx64.bashsh at gmail.com
>>>
>>> >> <mailto:linuxx64.bashsh at gmail.com>> wrote:
>>> >>
>>> >> Hi,
>>> >>
>>> >> First, excellent job! The custom Paul voice in the weather file
>>> is
>>> >> great. It has a lot of inflection and sounds like 4.4, which is
>>> pretty
>>> >> impressive. The test.wav is strange though. It's very buzzy and
>>> hard
>>> >> to understand. Just to clarify, for now, are we just testing the
>>> Paul
>>> >> voice?
>>> >>
>>> >> The weather sample seems to be very articulate and precise. It
>>> doesn't
>>> >> have any of that slurring of old versions on certain words. This
>>>
>>> is
>>> >> great.
>>> >>
>>> >> Just to hear the sample at a higher rate of speed, can you do
>>> the
>>> same
>>> >> text at about 350 wpm to see how it stands up?
>>> >>
>>> >> I didn't hear any of the soft tones at the end of sentences, and
>>> even
>>> >> ran the file and amplified silences, and it's about 48 dB below
>>> >> normal, so it's pretty quiet as it is without the code to squash
>>> the
>>> >> humming.
>>> >>
>>> >> Thanks,
>>> >>
>>> >> Alex
>>> >>
>>> >> On 8/29/11, ebruckert Bruckert <edbruckert at gmail.com
>>>
>>> >> <mailto:edbruckert at gmail.com>> wrote:
>>> >> > First some baseline testing stuff. Do not over test it in
>>> other
>>> >> words in the
>>> >> > beginning this is a rough cut so if we get into too many
>>> details
>>> >> or changes
>>> >> > it will just all become mud. We should only make a few changes
>>> >> at a time,
>>> >> > the other problem may become differences of opinion especially
>>> >> with some of
>>> >> > the voices which were never very stable because we never
>>> really
>>> >> liked
>>> >> > exactly what we had. This is an old version of DECtalk so
>>> >> occasionally you
>>> >> > will hear little homes at the end of things during what should
>>> >> be silence.
>>> >> > This is a mathematical problem called a limit cycle when we're
>>> >> finished I'll
>>> >> > add the new code I developed to squelch it. It took us years
>>> to
>>> >> figure out a
>>> >> > way to get rid of it perceptually we never could get rid of it
>>> >> > mathematically. Lastly I had to change the gain of voicing on
>>> >> the modified
>>> >> > voice that was sent without going into a long explanation
>>> which
>>> >> I can later
>>> >> > if people want the gains of the cascade vocal track have never
>>> >> ever been
>>> >> > constant it's a mathematical issue where you're constantly
>>> >> trying to adjust
>>> >> > them between quantization and overload and we've never been
>>> able
>>> to
>>> >> > guarantee them.
>>> >> > Lastly people can think about singing. The problem is that
>>> most
>>> SAPI
>>> >> > applications strip stuff out so I'm looking for suggestions of
>>> >> how we could
>>> >> > fool it and tell DECtalk that it was supposed to go into
>>> >> phonemic mode so it
>>> >> > can sing. I don't know what can and cannot get through the
>>> various
>>> >> > applications and while
>>> >> >
>>> >> > Again don't put too much effort in two critical listening at
>>> >> this stage as
>>> >> > were not fine-tuning it just training 3n where targets are
>>> >> >
>>> >> > On Mon, Aug 29, 2011 at 1:06 PM, jake mcmahan
>>>
>>> >> <mcmahan.jake at gmail.com <mailto:mcmahan.jake at gmail.com>>wrote:
>>>
>>> >> >
>>> >> >> On 8/29/2011 11:52 AM, ebruckert Bruckert wrote:
>>> >> >>
>>> >> >> Hello everyone the good news is I had a generator the bad
>>> news
>>> >> is I had to
>>> >> >> use it. I'm having trouble with the 2008 free edition of
>>> visual
>>> >> C I can't
>>> >> >> run the debugger. But before I uninstall and reinstall it I
>>> >> want to get
>>> >> >> this
>>> >> >> baseline out but it's a little bit painful because I can't
>>> run
>>> the
>>> >> >> debugger
>>> >> >> so I'm doing it by an archaic method. So it'll be out
>>> sometime
>>> >> today and
>>> >> >> then I'll reload visual C.
>>> >> >>
>>> >> >> On Sun, Aug 28, 2011 at 4:11 PM, jake mcmahan
>>>
>>> >> >> <mcmahan.jake at gmail.com
>>> <mailto:mcmahan.jake at gmail.com>>wrote:
>>>
>>> >> >>
>>> >> >> On 8/28/2011 11:40 AM, ebruckert Bruckert wrote:
>>> >> >>
>>> >> >> Ed K okay I'm ready. I have a base tuning of all the voices.
>>> So
>>> >> send me
>>> >> >> the
>>> >> >> various files to test with and we will try and get some
>>> >> agreement on the
>>> >> >> voices.
>>> >> >>
>>> >> >>
>>> >> >> On Sun, Aug 21, 2011 at 10:32 AM, ebruckert Bruckert
>>>
>>> >> <edbruckert at gmail.com <mailto:edbruckert at gmail.com>
>>>
>>> >> >> > wrote:
>>> >> >>
>>> >> >> I was visiting my grandson, back to woek today.....
>>> >> >>
>>> >> >>
>>> >> >> On Sat, Aug 20, 2011 at 11:45 PM, Alex H.
>>> >> >> <linuxx64.bashsh at gmail.com
>>>
>>> >> <mailto:linuxx64.bashsh at gmail.com>>wrote:
>>>
>>> >> >>
>>> >> >> Hi,
>>> >> >>
>>> >> >> Your message came through. DECTalk is alive and singing. :D
>>> >> >>
>>> >> >> Alex
>>> >> >> On 8/20/2011 11:22 PM, Doris and Chris wrote:
>>> >> >> > I have not seen any messages from this list in over a week.
>>> >> So I was
>>> >> >> > wondering if the list was down or if there is a problem at
>>> my
>>> >> end or
>>> >> >> > with my subscription.
>>> >> >> >
>>> >> >> > Cheers,
>>> >> >> >
>>> >> >> > Doris
>>> >> >> >
>>> >> >> > _______________________________________________
>>> >> >> > DECtalk mailing list
>>>
>>> >> >> > DECtalk at bluegrasspals.com
>>> <mailto:DECtalk at bluegrasspals.com>
>>>
>>> >> >> > http://bluegrasspals.com/mailman/listinfo/dectalk
>>> >> >>
>>> >> >> --
>>> >> >> --
>>> >> >> Sent via Thunderbird.
>>> >> >>
>>> >> >> _______________________________________________
>>> >> >> DECtalk mailing list
>>>
>>> >> >> DECtalk at bluegrasspals.com <mailto:DECtalk at bluegrasspals.com>
>>>
>>> >> >> http://bluegrasspals.com/mailman/listinfo/dectalk
>>> >> >>
>>> >> >>
>>> >> >>
>>> >> >>
>>> >> >>
>>> >> >> _______________________________________________
>>> >> >> DECtalk mailing
>>> >> >>
>>> >>
>>> >>
>>> listDECtalk at bluegrasspals.comhttp://bluegrasspals.com/mailman/listinfo/dectalk
>>>
>>> >> <http://bluegrasspals.com/mailman/listinfo/dectalk>
>>> >> >>
>>>
>>> >> >> take this, for example. This is a weather file.
>>> >> >>
>>> >> >> _______________________________________________
>>> >> >> DECtalk mailing list
>>>
>>> >> >> DECtalk at bluegrasspals.com <mailto:DECtalk at bluegrasspals.com>
>>>
>>> >> >> http://bluegrasspals.com/mailman/listinfo/dectalk
>>> >> >>
>>> >> >>
>>> >> >>
>>> >> >>
>>> >> >> _______________________________________________
>>> >> >> DECtalk mailing
>>> >> >>
>>> >>
>>> >>
>>> listDECtalk at bluegrasspals.comhttp://bluegrasspals.com/mailman/listinfo/dectalk
>>>
>>> >> <http://bluegrasspals.com/mailman/listinfo/dectalk>
>>> >> >>
>>>
>>> >> >> I gotcha man. If anyone can do it, you can dude
>>> >> >>
>>> >> >> _______________________________________________
>>> >> >> DECtalk mailing list
>>>
>>> >> >> DECtalk at bluegrasspals.com <mailto:DECtalk at bluegrasspals.com>
>>>
>>> >> >> http://bluegrasspals.com/mailman/listinfo/dectalk
>>> >> >>
>>> >> >>
>>> >> >
>>> >> _______________________________________________
>>> >> DECtalk mailing list
>>>
>>> >> DECtalk at bluegrasspals.com <mailto:DECtalk at bluegrasspals.com>
>>>
>>> >> http://bluegrasspals.com/mailman/listinfo/dectalk
>>> >>
>>> >>
>>> >>
>>> >>
>>> >> _______________________________________________
>>> >> DECtalk mailing list
>>> >> DECtalk at bluegrasspals.com
>>> >> http://bluegrasspals.com/mailman/listinfo/dectalk
>>> > again, nice job ed
>>> >
>>>
>>> _______________________________________________
>>> DECtalk mailing list
>>> DECtalk at bluegrasspals.com
>>> http://bluegrasspals.com/mailman/listinfo/dectalk
>>>
>>>
>>>
>>>
>>>
>>> --------------------------------------------------------------------------------
>>>
>>>
>>> _______________________________________________
>>> DECtalk mailing list
>>> DECtalk at bluegrasspals.com
>>> http://bluegrasspals.com/mailman/listinfo/dectalk
>>>
>> _______________________________________________
>> DECtalk mailing list
>> DECtalk at bluegrasspals.com
>> http://bluegrasspals.com/mailman/listinfo/dectalk
>>
>>
>
> _______________________________________________
> DECtalk mailing list
> DECtalk at bluegrasspals.com
> http://bluegrasspals.com/mailman/listinfo/dectalk
>
More information about the Dectalk
mailing list