adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?


Vy pham
 

Hi all,
Just wondering if there a way to purchase and add the Microsoft Azure
Neural Text-to-Speech to NVDA for personal use? I love the 2 voices
that are available for Vietnamese language, really high quality. I
checked out there site
https://azure.microsoft.com/en-us/services/cognitive-services/text-to-speech/?ocid=AID3027325.
Based on the descriptions, these voices are for Business Enterprise
usage, at least I am not so certain. Had anyone ever tried and had
any experience adding these voices to your screen readers? Are you all
aware of anyway to buy these voices as individual licenses? Similar to
Vocalizer? Any feedback would be very helpful.
Thanks!


Ali Savas
 

Hi,

In theory, it would certainly be possible to use Microsoft Azure voices
with an addon, however I'm afraid it doesn't make sense at all. There
are several reasons for this:

1. these voices are not installable, but always run online. You would
have to make sure you always have an internet connection.
2. let's assume that you have a permanent internet connection. Even
then, depending on your internet connection, you would have a very high
delay compared to the locally installed voices. Even if you had a very
good internet connection, the response would most likely still not be
high enough to work quickly and efficiently. Everything the voice is
supposed to say is calculated in the cloud and then played back. These
high quality voices that are processed in the cloud were not designed
for such deployments.
3. You can't just buy the Azure voices and use them at will. In the
Azure cloud, everything is charged per performance, compute cores,
queries, word count, etc. So depending on how you work it could get
pretty expensive.

I hope I was able to explain it in an understandable way. If something
is still unclear, just ask again.

Best regards
Ali


Sarah k Alawami
 

We're working on adding the neural tts voices to the program we use for flight simming. You can get an API key and it is something like 1mb per month for the neural tts voices, which given the amount of info we use is not going to matter, and we will have a fallback to sapi when internet falls over. Well I can't code but I am taking part in the discussion and implementation of the ideas.

-----Original Message-----
From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of Ali Savas
Sent: Saturday, June 5, 2021 5:07 AM
To: nvda@nvda.groups.io
Subject: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

Hi,

In theory, it would certainly be possible to use Microsoft Azure voices with an addon, however I'm afraid it doesn't make sense at all. There are several reasons for this:

1. these voices are not installable, but always run online. You would have to make sure you always have an internet connection.
2. let's assume that you have a permanent internet connection. Even then, depending on your internet connection, you would have a very high delay compared to the locally installed voices. Even if you had a very good internet connection, the response would most likely still not be high enough to work quickly and efficiently. Everything the voice is supposed to say is calculated in the cloud and then played back. These high quality voices that are processed in the cloud were not designed for such deployments.
3. You can't just buy the Azure voices and use them at will. In the Azure cloud, everything is charged per performance, compute cores, queries, word count, etc. So depending on how you work it could get pretty expensive.

I hope I was able to explain it in an understandable way. If something is still unclear, just ask again.

Best regards
Ali


mattias
 

Why are You so negative ali?

In chrome os You can for example use the internet based Android voices

Yes litle slow but functional

So why not the ms neural ones?

I’m litle thired about all this negativenes

 

 

Skickades från E-post för Windows 10

 

Från: Sarah k Alawami
Skickat: den 5 juni 2021 16:05
Till: nvda@nvda.groups.io
Ämne: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

 

We're working on adding the neural tts voices to the program we use for flight simming. You can get an API key and it is something like 1mb per month for the neural tts voices, which given the amount of info we use is not going to matter, and we will have a fallback to sapi when internet falls over. Well I can't code but I am taking part in the discussion and implementation of the ideas.

 

-----Original Message-----

From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of Ali Savas

Sent: Saturday, June 5, 2021 5:07 AM

To: nvda@nvda.groups.io

Subject: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

 

Hi,

 

In theory, it would certainly be possible to use Microsoft Azure voices with an addon, however I'm afraid it doesn't make sense at all. There are several reasons for this:

 

1. these voices are not installable, but always run online. You would have to make sure you always have an internet connection.

2. let's assume that you have a permanent internet connection. Even then, depending on your internet connection, you would have a very high delay compared to the locally installed voices. Even if you had a very good internet connection, the response would most likely still not be high enough to work quickly and efficiently. Everything the voice is supposed to say is calculated in the cloud and then played back. These high quality voices that are processed in the cloud were not designed for such deployments.

3. You can't just buy the Azure voices and use them at will. In the Azure cloud, everything is charged per performance, compute cores, queries, word count, etc. So depending on how you work it could get pretty expensive.

 

I hope I was able to explain it in an understandable way. If something is still unclear, just ask again.

 

Best regards

Ali

 

 

 

 

 

 

 

 

 

 

 


 

On Sun, Jun 6, 2021 at 10:50 AM, mattias wrote:
I’m litle thired about all this negativenes
-
Sorry, but there is no requirement that any member be a cheerleader for anything.

Those "not so hot" on the idea have given very clear reasons why they're not.  That's entirely appropriate.

Feedback, whether positive or negative, that's well-explained is very welcome here.
--

Brian - Windows 10, 64-Bit, Version 21H1, Build 19043  

I do not understand why some seek to separate a person from their actions.  The self is composed of an individual’s thoughts, actions, and expression, which are contained in and actuated by the body.  What you do and say is the clearest indicator of who you are.

      ~ Brian Vogel

 


zvonimir stanečić, 9a5dsz
 

That’s your right.

Not all people have the access to the high speed internet to use this in the functional manner.

There was a Hebrew tts, called kolfix Sivan. It was cloud based, sapi5 compliant. It worked very slow, and the response was not so good, because of the cloud synthesis processing.

Best,

Zvonimir

 

From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of mattias
Sent: Sunday, June 6, 2021 8:55 AM
To: nvda@nvda.groups.io
Subject: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

 

Why are You so negative ali?

In chrome os You can for example use the internet based Android voices

Yes litle slow but functional

So why not the ms neural ones?

I’m litle thired about all this negativenes

 

 

Skickades från E-post för Windows 10

 

Från: Sarah k Alawami
Skickat: den 5 juni 2021 16:05
Till: nvda@nvda.groups.io
Ämne: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

 

We're working on adding the neural tts voices to the program we use for flight simming. You can get an API key and it is something like 1mb per month for the neural tts voices, which given the amount of info we use is not going to matter, and we will have a fallback to sapi when internet falls over. Well I can't code but I am taking part in the discussion and implementation of the ideas.

 

-----Original Message-----

From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of Ali Savas

Sent: Saturday, June 5, 2021 5:07 AM

To: nvda@nvda.groups.io

Subject: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

 

Hi,

 

In theory, it would certainly be possible to use Microsoft Azure voices with an addon, however I'm afraid it doesn't make sense at all. There are several reasons for this:

 

1. these voices are not installable, but always run online. You would have to make sure you always have an internet connection.

2. let's assume that you have a permanent internet connection. Even then, depending on your internet connection, you would have a very high delay compared to the locally installed voices. Even if you had a very good internet connection, the response would most likely still not be high enough to work quickly and efficiently. Everything the voice is supposed to say is calculated in the cloud and then played back. These high quality voices that are processed in the cloud were not designed for such deployments.

3. You can't just buy the Azure voices and use them at will. In the Azure cloud, everything is charged per performance, compute cores, queries, word count, etc. So depending on how you work it could get pretty expensive.

 

I hope I was able to explain it in an understandable way. If something is still unclear, just ask again.

 

Best regards

Ali

 

 

 

 

 

 

 

 

 

 

 


Sarah k Alawami
 

Ping rates etc, can affect the voices. If you have internet breakage, there goes your speech. Also, if your internet connection brakes or burps in the middle of a sentence, there goes your speech, plus I believe that API key is expensive and you’re limited to 1mb probably a month for the free account.

 

From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of mattias
Sent: Saturday, June 5, 2021 11:55 PM
To: nvda@nvda.groups.io
Subject: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

 

Why are You so negative ali?

In chrome os You can for example use the internet based Android voices

Yes litle slow but functional

So why not the ms neural ones?

I’m litle thired about all this negativenes

 

 

Skickades från E-post för Windows 10

 

Från: Sarah k Alawami
Skickat: den 5 juni 2021 16:05
Till: nvda@nvda.groups.io
Ämne: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

 

We're working on adding the neural tts voices to the program we use for flight simming. You can get an API key and it is something like 1mb per month for the neural tts voices, which given the amount of info we use is not going to matter, and we will have a fallback to sapi when internet falls over. Well I can't code but I am taking part in the discussion and implementation of the ideas.

 

-----Original Message-----

From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of Ali Savas

Sent: Saturday, June 5, 2021 5:07 AM

To: nvda@nvda.groups.io

Subject: Re: [nvda] adding Microsoft Neural Text-to-Speech to NVDA? Is that possible?

 

Hi,

 

In theory, it would certainly be possible to use Microsoft Azure voices with an addon, however I'm afraid it doesn't make sense at all. There are several reasons for this:

 

1. these voices are not installable, but always run online. You would have to make sure you always have an internet connection.

2. let's assume that you have a permanent internet connection. Even then, depending on your internet connection, you would have a very high delay compared to the locally installed voices. Even if you had a very good internet connection, the response would most likely still not be high enough to work quickly and efficiently. Everything the voice is supposed to say is calculated in the cloud and then played back. These high quality voices that are processed in the cloud were not designed for such deployments.

3. You can't just buy the Azure voices and use them at will. In the Azure cloud, everything is charged per performance, compute cores, queries, word count, etc. So depending on how you work it could get pretty expensive.

 

I hope I was able to explain it in an understandable way. If something is still unclear, just ask again.

 

Best regards

Ali

 

 

 

 

 

 

 

 

 

 

 


Ali Savas
 

Hi Mattias,

Why are You so negative ali?
I am anything but negative. On the contrary, I am one of those people
who are very keen to experiment and try out a lot of things. You can
call me a very curious person. I just wanted to try to explain
technically why such an addon might not be an option for some users or
why it might not be useful.

There was even an issue on Github. Since these voices require one key
per user and setting this up might be a challenge for some users, the
developers have rejected it and left it up to the addon developers to
decide if there will be an addon for this or not.

If there will be an addon at some point, I will create an Azure account,
leave my payment details with Microsoft and just give it a try.

Best regards
Ali