Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

Hi,
Thanks for letting me know how the format shows up for you.
Yea, I've heard that Alpine is good. Is it only for the Console? Can it be used in Graphical too? I'm mostly a Graphical user, but even if its only for Console, I may still check it out.
Yea, I agree that its a good idea to have a wide variety of voices to choose from. Its all a matter of personal preference and I'd like for Linux users to have more of those choices.
Personally I'm not a fan of the more computerized sounding voices. I much prefer the more natural sounding voices if I have the choice. I'd like for there to be as many natural voices for Linux as there are computerized ones. Voxin with Nuance is a good one for natural voices. But I'd like to see more diversity of different TTS synths, every synth has its own style.
And for those of us who prefer natural voices, Voxin is basically our only choice for Linux. It would be good to have some other choices there. If this project moves forward, it would be great to have Readspeaker as another option.
Anyways, thanks for letting me know about the formatting, and for your input re TTS.
SL

"The original formatting was fine here too.  I suspect the email client and its configuration for text handling is a factor.  I'm using alpine.

>> For what it's worth, I didn't notice anything off with the original
>> message's formatting...
>>
>> As for the question, I'm quite content with espeak-ng's default voice
>> and consider it higher quality than the more natural sounding voices
>> I've heard, many of which I feel fall into the Uncanny Valley if used
>> to read anything longer than a single sentence...
>>
>> That said, more choice is generally a good thing."
>>
>> _______________________________________________
>> Blinux-list mailing list
>> Blinux-list at redhat.com
>> https://listman.redhat.com/mailman/listinfo/blinux-list
>>
>
> -- 
> XR
>
> _______________________________________________
> Blinux-list mailing list
> Blinux-list at redhat.com
> https://listman.redhat.com/mailman/listinfo/blinux-list
>

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

Don't get me wrong, more natural sounding TTS with proper inflection
would be great, and for me, the holy grail would be TTS capable of
reading a digitized novel in real-time or reading subtitles on foreign
media in real-time and be indistinguishable from a human cast
recording a audio dramatization or dubbed vocal track... but unless
there's been massive improvements in recent years I'm unaware of, the
natural voices are at that point where they almost sound human but
fail in a subtle but unsettling way that's hard to qualify, and until
we get over that hurdle, I'll take the obviously robotic monotone over
the almost, but not quite, passes for a human reader voices for daily
work.

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

The problem is that all the so called human voices are spliced together syllables and word fragments taped together. So you get emphasis on the wrong parts of the sentences, pauses in the wrong place, and stuff like that. If they would devote more machine learning time into proper text to speech rendering instead of sensorship and other nonsense, we might get somewhere.

----- Original Message -----
From: Linux for blind general discussion <blinux-list at redhat.com>
To: blinux-list at redhat.com
Date: Sun, 18 Apr 2021 00:42:25 +0000
Subject: Re: Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

> Don't get me wrong, more natural sounding TTS with proper inflection
> would be great, and for me, the holy grail would be TTS capable of
> reading a digitized novel in real-time or reading subtitles on foreign
> media in real-time and be indistinguishable from a human cast
> recording a audio dramatization or dubbed vocal track... but unless
> there's been massive improvements in recent years I'm unaware of, the
> natural voices are at that point where they almost sound human but
> fail in a subtle but unsettling way that's hard to qualify, and until
> we get over that hurdle, I'll take the obviously robotic monotone over
> the almost, but not quite, passes for a human reader voices for daily
> work.
> 
> _______________________________________________
> Blinux-list mailing list
> Blinux-list at redhat.com
> https://listman.redhat.com/mailman/listinfo/blinux-list
> 
> 

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

I'll admit, I haven't paid much attention to machine learning... but
then again, I get the impression that the training phase generally
requires a server farm worth of resources if you want to build
something that actually out performs oldschool algorithmic AI, and
that even once trained, you generally can't just export the end result
so it'll run on an end-user's machine... and I, for one, would hate to
have my TTS cut out just because my Internet connection glitches out.

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

Yeah, sometimes the splicing with natural-like voices can be quite irritating. I'd still rather that than computerized voices though lol.
? I think Ivona did pretty good at making almost human-like voices. I havn't found any other TTS that compares to Ivona. Like, if you ever have the chance to listen to the Ivona voice called Kendra, of course, its not completely human, but it has some of the most natural flow and accurate inflection I've ever heard in a TTS. I have Ivona's Kendra on Windows with NVDA and I just increase the speed a little bit (still at normal human speaking speed just a bit faster than the default rate) and to me, that sounds even more natural than the default speed used on Bookshare for example. Last I checked, Bookshare used Ivona Kendra to voice its books when you choose audio. 
But when I listen to something like the Nuance voices, the inflection just doesn't compare to the flow of Ivona. Thats why its so frustrating that we can't seem to get Ivona on Linux.
Still, I would? take Nuance or just about any other natural TTS synth, even if it has some aukward inflections, I'd still take that any day, over Espeak or Eloquence, but thats just me lol. I can tollerate short tasks with computerized speech, but if its a longer piece of text, anything that I am supposed to focus on or enjoy, then I want a natural TTS for that. But I'm glad the computerized voices are available. Its really just a personal preference. Some like the computerized ones, and some like the natural-ish ones. I've always chose the natural ones whenever I have the choice. Even though they're not quite human, to me I just find them easier to listen to.
SL

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

But this is one thing I find confusing...at least for Linux.
tts is not a screen reader program.
One may incorporate a tts  module into the workings of a screen reader 
program, the way drivers were written to allow hardware synthesizers to 
communicate with said program, but the tts itself is not going to, on its 
own, manage things like responsiveness while typing and the like...and 
that is before you talk of latency problem possibilities.
It is the screen reader program itself that, in my experience, takes care 
of inflection, allowing the user to get more or less, same thing with 
punctuation marks, pitch and speed.
If speak reader is strictly a tts, the company   may not understand the 
need for  things like making sure the tts can follow activity and control 
of the  computer itself.
  before  writing this email I did a quick google  using the phrase tts 
defined?
with  the first several options discussing how those with reading 
challenges   like dyslexia use tts to manage small blocks of words on the 
screen  with the recommended rate of 180 words per minute..or less.
It is, speaking personally, very unfortunate that some think a tts is a 
screen reader program, when in reality they are different.
I have a friend who likes to use her amazon kindle to read fanfiction 
aloud.
We have these discussions because my screen reader has  no issue  properly 
pronouncing say  the name of Ron Weasley from the harry Potter books, but 
the Kindle tts cannot pronounce the word correctly at all.
Do not be surprised if you end up needing to demonstrate how your screen 
reader, orca or speakup, does more than just read text, which for many  is 
the only   purpose of a tts    tool.
Does that make sense?



On Sat, 17 Apr 2021, Linux for blind general discussion wrote:

> The problem is that all the so called human voices are spliced together syllables and word fragments taped together. So you get emphasis on the wrong parts of the sentences, pauses in the wrong place, and stuff like that. If they would devote more machine learning time into proper text to speech rendering instead of sensorship and other nonsense, we might get somewhere.
>
> ----- Original Message -----
> From: Linux for blind general discussion <blinux-list at redhat.com>
> To: blinux-list at redhat.com
> Date: Sun, 18 Apr 2021 00:42:25 +0000
> Subject: Re: Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included
>
>> Don't get me wrong, more natural sounding TTS with proper inflection
>> would be great, and for me, the holy grail would be TTS capable of
>> reading a digitized novel in real-time or reading subtitles on foreign
>> media in real-time and be indistinguishable from a human cast
>> recording a audio dramatization or dubbed vocal track... but unless
>> there's been massive improvements in recent years I'm unaware of, the
>> natural voices are at that point where they almost sound human but
>> fail in a subtle but unsettling way that's hard to qualify, and until
>> we get over that hurdle, I'll take the obviously robotic monotone over
>> the almost, but not quite, passes for a human reader voices for daily
>> work.
>>
>> _______________________________________________
>> Blinux-list mailing list
>> Blinux-list at redhat.com
>> https://listman.redhat.com/mailman/listinfo/blinux-list
>>
>>
>
>
> _______________________________________________
> Blinux-list mailing list
> Blinux-list at redhat.com
> https://listman.redhat.com/mailman/listinfo/blinux-list
>
>

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

My answer to this repetitive question is most definitely no and for the 
same reason speech synthesizers weren't originally given naturally 
sounding voices in the beginning of their development.  Speech 
synthesizers in the early 1960's were top secret equipment and put in 
military fighter aircraft and maybe also bombers.  The reason they didn't 
get natural sounding voices then was that the air crews needed to be able 
to distinguish speech synthesizer announcements from other natural 
sounding speech from over the intercom and over the radio.  Having lived 
with synthetic sounding speech in my case since 1987 not only am I used to 
it, as a result of research I did on its origins I understand its purpose 
and proper use.  Can synthetic speech be left synthetic and get around 
people's auditory difficulties?  That I don't know but that could be a 
helpful line of research.



On Sun, 18 Apr 2021, Linux for blind general discussion wrote:

> But this is one thing I find confusing...at least for Linux.
> tts is not a screen reader program.
> One may incorporate a tts  module into the workings of a screen reader 
> program, the way drivers were written to allow hardware synthesizers to 
> communicate with said program, but the tts itself is not going to, on its 
> own, manage things like responsiveness while typing and the like...and that 
> is before you talk of latency problem possibilities.
> It is the screen reader program itself that, in my experience, takes care of 
> inflection, allowing the user to get more or less, same thing with 
> punctuation marks, pitch and speed.
> If speak reader is strictly a tts, the company   may not understand the need 
> for  things like making sure the tts can follow activity and control of the 
> computer itself.
> before  writing this email I did a quick google  using the phrase tts 
> defined?
> with  the first several options discussing how those with reading challenges 
> like dyslexia use tts to manage small blocks of words on the screen  with the 
> recommended rate of 180 words per minute..or less.
> It is, speaking personally, very unfortunate that some think a tts is a 
> screen reader program, when in reality they are different.
> I have a friend who likes to use her amazon kindle to read fanfiction aloud.
> We have these discussions because my screen reader has  no issue  properly 
> pronouncing say  the name of Ron Weasley from the harry Potter books, but the 
> Kindle tts cannot pronounce the word correctly at all.
> Do not be surprised if you end up needing to demonstrate how your screen 
> reader, orca or speakup, does more than just read text, which for many  is 
> the only   purpose of a tts    tool.
> Does that make sense?
>
>
>
> On Sat, 17 Apr 2021, Linux for blind general discussion wrote:
>
>>  The problem is that all the so called human voices are spliced together
>>  syllables and word fragments taped together. So you get emphasis on the
>>  wrong parts of the sentences, pauses in the wrong place, and stuff like
>>  that. If they would devote more machine learning time into proper text to
>>  speech rendering instead of sensorship and other nonsense, we might get
>>  somewhere.
>>
>>  ----- Original Message -----
>>  From: Linux for blind general discussion <blinux-list at redhat.com>
>>  To: blinux-list at redhat.com
>>  Date: Sun, 18 Apr 2021 00:42:25 +0000
>>  Subject: Re: Formatting - was Would you be interested in having natural
>>  sounding TTS voices by Readspeaker on Linux? demo link included
>>
>>>  Don't get me wrong, more natural sounding TTS with proper inflection
>>>  would be great, and for me, the holy grail would be TTS capable of
>>>  reading a digitized novel in real-time or reading subtitles on foreign
>>>  media in real-time and be indistinguishable from a human cast
>>>  recording a audio dramatization or dubbed vocal track... but unless
>>>  there's been massive improvements in recent years I'm unaware of, the
>>>  natural voices are at that point where they almost sound human but
>>>  fail in a subtle but unsettling way that's hard to qualify, and until
>>>  we get over that hurdle, I'll take the obviously robotic monotone over
>>>  the almost, but not quite, passes for a human reader voices for daily
>>>  work.
>>>
>>>  _______________________________________________
>>>  Blinux-list mailing list
>>>  Blinux-list at redhat.com
>>>  https://listman.redhat.com/mailman/listinfo/blinux-list
>>> 
>>> 
>> 
>>
>>  _______________________________________________
>>  Blinux-list mailing list
>>  Blinux-list at redhat.com
>>  https://listman.redhat.com/mailman/listinfo/blinux-list
>> 
>> 
>
> _______________________________________________
> Blinux-list mailing list
> Blinux-list at redhat.com
> https://listman.redhat.com/mailman/listinfo/blinux-list
>
>
>

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

Yes, makes sense.
I think how it works in many cases, is that there is the screenreader program, the TTS software, and then a module that kind of links the two together to function.
So like, I'll use Windows here as an example just because thats what I've used most over the years until recently getting into Linux. But on Windows, you can have the NVDA screenreader, then you can buy an Ivona TTS voice. But then, if you want to use the Ivona voice through NVDA, you have to select the, I think its the SAPI 5 synth if I remember correctly? You need to select that SAPI 5 synth in the NVDA settings in order to get the Ivona TTS voice to work with NVDA, and then you control the pitch and pronounciation etc of Ivona through NVDA.
 I think its kinda similar with the Orca screenreader in Linux. So if you're using the RHVoice voices, you have to select for it to use the RHVoice synth module so they all work together. As far as I'm aware, every TTS solution needs a module of sorts to connect it with the screenreader.
one of the guys either on the Readspeaker team, or associated with the team, is blind. I am not communicating with him directly yet, but the representative I am in contact with mentioned him. It would be great if he might consider collaborating? in the project as I am assuming he probably uses a screenreader of some sort. Even if he primarily uses Braille, (which I don't know that for a fact, I'm just mentioning it as a possibility), but even if he did mostly use braille, the fact that he is involved with a company that manufactures TTS voice solutions for desktop, servers, embedded systems etc, means he likely has quite a bit of knowledge regarding the technicalities of TTS and screenreaders, and if he wanted to collaborate with the project, he would likely be helpful in advising them there on the ground if needed. I can communicate with them remotely, but its always really great to have someone that can sit there in person with the developers. I don't know if he works right there with them, or remotely. But I'll ask when I next contact them on a week day.
Thanks,
SL

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

Actually, that is not the nature of my question at all.
I would add that some  aspects of Speech synthesis was created for those 
who,  say due to military service, lost the ability to speak.
Still, my question here refers to the idea, which you seem to echo, that 
tts, software, and screen reader software are interchangeable.
They are not, which is why I was surprised that some here feel speak reader 
will perform the function of a screen reader, managing response to say 
typing, and inflection...which should be controlled by the individual user, 
speaking personally.




On Sun, 18 Apr 2021, Linux for blind general discussion wrote:

> My answer to this repetitive question is most definitely no and for the same 
> reason speech synthesizers weren't originally given naturally sounding voices 
> in the beginning of their development.  Speech synthesizers in the early 
> 1960's were top secret equipment and put in military fighter aircraft and 
> maybe also bombers.  The reason they didn't get natural sounding voices then 
> was that the air crews needed to be able to distinguish speech synthesizer 
> announcements from other natural sounding speech from over the intercom and 
> over the radio.  Having lived with synthetic sounding speech in my case since 
> 1987 not only am I used to it, as a result of research I did on its origins I 
> understand its purpose and proper use.  Can synthetic speech be left 
> synthetic and get around people's auditory difficulties?  That I don't know 
> but that could be a helpful line of research.
>
>
>
> On Sun, 18 Apr 2021, Linux for blind general discussion wrote:
>
>>  But this is one thing I find confusing...at least for Linux.
>>  tts is not a screen reader program.
>>  One may incorporate a tts  module into the workings of a screen reader
>>  program, the way drivers were written to allow hardware synthesizers to
>>  communicate with said program, but the tts itself is not going to, on its
>>  own, manage things like responsiveness while typing and the like...and
>>  that is before you talk of latency problem possibilities.
>>  It is the screen reader program itself that, in my experience, takes care
>>  of inflection, allowing the user to get more or less, same thing with
>>  punctuation marks, pitch and speed.
>>  If speak reader is strictly a tts, the company   may not understand the
>>  need for  things like making sure the tts can follow activity and control
>>  of the computer itself.
>>  before  writing this email I did a quick google  using the phrase tts
>>  defined?
>>  with  the first several options discussing how those with reading
>>  challenges like dyslexia use tts to manage small blocks of words on the
>>  screen  with the recommended rate of 180 words per minute..or less.
>>  It is, speaking personally, very unfortunate that some think a tts is a
>>  screen reader program, when in reality they are different.
>>  I have a friend who likes to use her amazon kindle to read fanfiction
>>  aloud.
>>  We have these discussions because my screen reader has  no issue  properly
>>  pronouncing say  the name of Ron Weasley from the harry Potter books, but
>>  the Kindle tts cannot pronounce the word correctly at all.
>>  Do not be surprised if you end up needing to demonstrate how your screen
>>  reader, orca or speakup, does more than just read text, which for many  is
>>  the only   purpose of a tts    tool.
>>  Does that make sense?
>> 
>> 
>>
>>  On Sat, 17 Apr 2021, Linux for blind general discussion wrote:
>> 
>> >   The problem is that all the so called human voices are spliced together
>> >   syllables and word fragments taped together. So you get emphasis on the
>> >   wrong parts of the sentences, pauses in the wrong place, and stuff like
>> >   that. If they would devote more machine learning time into proper text 
>> >   to
>> >   speech rendering instead of sensorship and other nonsense, we might get
>> >   somewhere.
>> > 
>> >   ----- Original Message -----
>> >   From: Linux for blind general discussion <blinux-list at redhat.com>
>> >   To: blinux-list at redhat.com
>> >   Date: Sun, 18 Apr 2021 00:42:25 +0000
>> >   Subject: Re: Formatting - was Would you be interested in having natural
>> >   sounding TTS voices by Readspeaker on Linux? demo link included
>> > 
>> > >   Don't get me wrong, more natural sounding TTS with proper inflection
>> > >   would be great, and for me, the holy grail would be TTS capable of
>> > >   reading a digitized novel in real-time or reading subtitles on 
>> > >   foreign
>> > >   media in real-time and be indistinguishable from a human cast
>> > >   recording a audio dramatization or dubbed vocal track... but unless
>> > >   there's been massive improvements in recent years I'm unaware of, the
>> > >   natural voices are at that point where they almost sound human but
>> > >   fail in a subtle but unsettling way that's hard to qualify, and until
>> > >   we get over that hurdle, I'll take the obviously robotic monotone 
>> > >   over
>> > >   the almost, but not quite, passes for a human reader voices for daily
>> > >   work.
>> > > 
>> > >   _______________________________________________
>> > >   Blinux-list mailing list
>> > >   Blinux-list at redhat.com
>> > >   https://listman.redhat.com/mailman/listinfo/blinux-list
>> > > 
>> > > 
>> > 
>> > 
>> >   _______________________________________________
>> >   Blinux-list mailing list
>> >   Blinux-list at redhat.com
>> >   https://listman.redhat.com/mailman/listinfo/blinux-list
>> > 
>> > 
>>
>>  _______________________________________________
>>  Blinux-list mailing list
>>  Blinux-list at redhat.com
>>  https://listman.redhat.com/mailman/listinfo/blinux-list
>> 
>> 
>> 
>
> _______________________________________________
> Blinux-list mailing list
> Blinux-list at redhat.com
> https://listman.redhat.com/mailman/listinfo/blinux-list
>
>

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

I had always assumed early speech synthesizers where obviously
non-human because the technology simply didn't exist to make more
human sounding voices viable...

As for responsiveness, I think there are two different things being
conflated here. It is indeed erroneous to equate TTS with the frontend
tool using it, whether that frontend is a screen reader, a mechanical
voicebox for the mute, or what have you, and yes, the screen reader,
if that's what's using the TTS should be controlling most of what the
TTS is doing, but it's important that the TTS be able to render speech
quickly enough that the screen reader can actually use it.

A TTS that can take an eBook as input and spit out an audiobook
indistinguishable from one recorded by a professional reader in a
sound booth would be great for generating audiobooks even if it took
twice as long to render the audio as to play it back, but it would be
kind of lousy to use with a screen reader if it took 5 seconds to
speak everytime Orca sent it a sentence to speak. I could be wrong,
but I suspect this is the kind of thing whoever originally asked about
responsiveness was talking about.

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

My goal is simply to bring more variety of TTS voices which could be installed to use with the already existing Linux screenreaders. For example, in Windows I can use the NVDA screenreader to adjust the rate and pitch etc of a TTS voice. But when all is said and done, I prefer the sound quality, the tamber, of certain TTS voices more than others, no matter how much I can adjust perameters in the screenreader. On Windows, I can use the Microsoft Zira voice for example, or the Ivona Kendra voice. No matter how many adjustments I make from the screenreader, granted that I am an average user with no programming experience, but no matter what standard adjustments I make, for me, it still comes down to quality of voice, and I'll still choose Kendra over Zira if I have the chance. So, I just want to see more choices of TTS voices? become available to Linux users.
The adjustments are done through the screenreader yes, but I once tried out a third party TTS on Windows. It was some years back, but I think I used it with the NVDA screenreader, and unlike all the other TTS voices I have ever used with NVDA on Windows, the one produced by that specific company, always had a very long pause between sentences no matter what adjustments I made from within the NVDA screenreader. It was a trial demo, so I chose not to buy the TTS product, because no matter what adjustments I made in the screenreader, I couldn't get rid of the lagging pauses between sentences which made reading documents with that particular TTS intolerable. I have not experienced that with any of the other TTS voices from various different companies I have used. Only? voices from that specific company, and I suspect it may have had something to do with whatever module they developed to connect their voices to the screenreader.
I have also experienced TTS voices that make like a slight popping sound as it tries to keep up with the key strokes when writing fast or navigating fast, as if the output can't quite keep up with the input.
 I am not a tech specialist, but based on those kind of? experiences, it leads me to believe that responsiveness may not just be the screenreader itself, but also may depend on the seamless connectivity between the screenreader, and the TTS module that its connected to. I could be wrong. I don't know all the hows and whys, but those have just been some of my experiences.
not everyone will be interested in Readspeaker and thats totally fine. Some people are satisfied with whats already available and I respect their preferences.
My hope for this potential project, is simply to make available more ?natural-like voices compatible with Linux screenreaders for those who want them.
Not everyone will want natural-like voices and thats ok.
But if this project moves forward, then I do want to be certain that there will not be any lagging in response with the connection between the TTS module, and the screenreader. So if Readspeaker develops a module for their voices to link to Orca and SpeechedUp, then I do think that the fluid responsiveness between Orca and Espeak-ng, or Orca and RHVoice, is a good example of seamless responsiveness, for Readspeaker to refer to as a model when testing their own module to be sure that their Readspeaker TTS module flows smoothly in connection with the screenreader as the user navigates.
Thanks,
SL

Formatting - was Would you be interested in having natural sounding TTS voices by Readspeaker on Linux? demo link included

[blinux-list]

The original formatting was fine here too.  I suspect the email client and its configuration for text handling is a factor.  I'm using alpine.

> For what it's worth, I didn't notice anything off with the original
> message's formatting...
>
> As for the question, I'm quite content with espeak-ng's default voice
> and consider it higher quality than the more natural sounding voices
> I've heard, many of which I feel fall into the Uncanny Valley if used
> to read anything longer than a single sentence...
>
> That said, more choice is generally a good thing.
>
> _______________________________________________
> Blinux-list mailing list
> Blinux-list at redhat.com
> https://listman.redhat.com/mailman/listinfo/blinux-list
>
>

-- 
XR