[GUI Plugin] TTS to MP3: Create MP3 audiobook using Windows TTS

[jackie_w]

Originally Posted by b00kiest

<snip>It was a rude awakening that in windows you have nothing easily set to do TTS and then once you can get it figured out you are left with David and Hazel<snip>

In Windows 10 there should be many more free MS Voices available to you than just David & Hazel. These 2 possibly date from Windows 7 or 8.

The first thing to do is to install as many of the Win10 Language packs as you're interested in. Your top-right info panel doesn't include your Country but your posts read like you're an English speaker. If so, all these English-dialect Voices can be installed for free:

Extract from post #2 in this thread:

Voices marked with * are Cortana Speech Assistant voices.
English - GB: Hazel, Susan, George, Sarah*
English - US: Zira, David, Mark, Eva*
English - Canada: Linda, Richard, Eva*
English - Australia: Catherine, James, Matilda*
English - India: Heera, Ravi, Koyal*
English - Ireland: Sean
N.B. Cortana Eva (US) and Cortana Eva (Canada) are stored as separate files so don't assume they are the same thing.

For detailed instructions, see next 2 posts:

• Installing extra MS Windows 10 Language packs
• Making sure already-installed MS Voices are also available within calibre

Please let me know how you get on especially if the instructions below need clarification.

[jackie_w]

Install new MS Win10 Language-dialect Voice packs via standard Win10
Settings > Time & Language > Speech > Manage voices > [+]Add voices

This will add physical Voice data files in subdirectories below this directory:

C:\Windows\Speech_OneCore\Engines\TTS\

All Voices for a given Language-dialect (e.g. en-GB, en-US etc.) are stored in the same subdirectory. Each MS Voice consists of several physical files.

[jackie_w]

1. Installation of a Win10 Language pack should also result in all Voices in the pack being automatically added as sub-keys below this Win10 registry key:
   

   HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech_OneCore\Voices\Tokens

   Use the Windows regedit utility to drill down to the above key and expand it to check the full list of installed Voices on your PC.
   
2. MS Voices available to calibre are stored under a different Windows registry key. The key depends on whether you're using calibre 32-bit or calibre 64-bit:
   • Calibre for Windows 64-bit:
     

     HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech\Voices\Tokens

   • Calibre for Windows 32-bit (including Calibre Portable):
     

     HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\SPEECH\Voices\Tokens

   So, if there are MS Voices in the list from step 1. which do not appear in the list from step 2. we need to find a way of duplicating the registry Voice data to the relevant registry key where your version of calibre will see it.
   ***This is not difficult but it is Windows registry hacking, so proceed at your own risk***
3. Close calibre before you start.
4. Using the en-GB Cortana Voice, named Sarah, as a worked example:
   Use regedit to drill down to the key shown in item 1. Expand this key to see its Voice list and right-click on the missing Voice's sub-key, in this case
   

   MSTTS_V110_enGB_SarahM

5. You will be offered a pop-up sub-menu. Select Export. Then save the file to somewhere you can find it again on your PC. It will have file extension .reg
   Please be aware that Windows .reg files behave like Windows .bat files if you double-click them, i.e. they run, so DON'T DO THAT - at least not yet!
6. Find the saved .reg file and open it in your favourite text editor (I use Notepad++ on Windows)
   See screenshot below. You need to manually edit the 2 yellow highlighted rows containing registry keys
   

   [HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech_OneCore\Voices\Tokens\MSTTS_V110_enGB_SarahM]
   [HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech_OneCore\Voices\Tokens\MSTTS_V110_enGB_SarahM\Attributes]

   Depending whether you're using calibre 64-bit or calibre 32-bit make the following manual changes:
   • Calibre for Windows 64-bit:
     Change both occurrences of
     

     \Microsoft\Speech_OneCore\

     to
     

     \Microsoft\Speech\

   • Calibre for Windows 32-bit (including Calibre Portable):
     Change both occurrences of
     

     \Microsoft\Speech_OneCore\

     to
     

     \WOW6432Node\Microsoft\SPEECH\

7. Save the updated .reg file. Run it by double-clicking on it. Windows will present you with a couple of "are you really sure you want to do this" confirmation pop-ups which you'll need to OK.
8. Launch calibre again and open this plugin (or the calibre Viewer). If all has gone to plan, you should now see MS Cortana Voice en-GB Sarah in the Voices drop-down list.

Repeat steps 3-8 for any installed MS Voice currently "missing" from calibre.

Notes:

• Don't be surprised if one day one of your MS automatic Win10 updates "helpfully" removes the calibre-visible registry entries you've just added. In other words, save any edited .reg files you've just created for when you might need them again. Not all MS updates do this but I've had at least one which did!
• If you purchase any 3rd-party Voices their installation file should automatically add the registry keys in the correct place. For example, I own 3 Ivona Voices. All 3 are automatically visible to both calibre 32-bit and calibre 64-bit without any registry hacking being necessary.
• Some purchased 3rd-party Voices may only be available to calibre 32-bit. If you're normally a calibre 64-bit user you can still access them by installing and using a separate copy of Calibre Portable which is a 32-bit application.

[b00kiest]

Originally Posted by jackie_w

In Windows 10 there should be many more free MS Voices available to you than just David & Hazel. These 2 possibly date from Windows 7 or 8.

The first thing to do is to install as many of the Win10 Language packs as you're interested in. Your top-right info panel doesn't include your Country but your posts read like you're an English speaker. If so, all these English-dialect Voices can be installed for free:


For detailed instructions, see next 2 posts:

• Installing extra MS Windows 10 Language packs
• Making sure already-installed MS Voices are also available within calibre

Please let me know how you get on especially if the instructions below need clarification.

Ok so silly me, I found some of them but none of the Cortana ones. Still don't sound as natural as my Mac but I'm not overly picky. I mean I have been listening to Susan and David pre Mac for years doing tts and let me just say they didn't enunciate well in the early days

Anywho, How do you get the Cortana ones. I may have turned them off as I don't use Cortana for anything.

If I didn't say this before many thanks! This will make my books more enjoyable and keep the pup from boredom as I work from home.

[b00kiest]

Originally Posted by jackie_w

Install new MS Win10 Language-dialect Voice packs via standard Win10
Settings > Time & Language > Speech > Manage voices > [+]Add voices

This will add physical Voice data files in subdirectories below this directory:

C:\Windows\Speech_OneCore\Engines\TTS\

All Voices for a given Language-dialect (e.g. en-GB, en-US etc.) are stored in the same subdirectory. Each MS Voice consists of several physical files.

Interestinly, I can see the Cortana Voices installed in this folder but they are not listed in the TTS area.

[b00kiest]

YOU ARE BrILLIANT. Now it's at least decent amount of Voices to play with. I would argue the quality compared to the ones I had on the Mac pale in comparison in that few of them can I hear the computerness but it's very doable. When I have some time I'll play with the speed and see if there are pitch change options anywhere to see if I can even out the robotiness in some of the voices. Could be the laptop sound card was not the upgrade if I recall I needed graphics more than I care about sound :/

Anyway, it worked like a charm. I even just exported at the token level to get them all in one file to find and replace more easily than to do it for each voice. So i'll keep the two files in case any updates overwrite this brilliance.

Again and always. THANK YOU!
Bookiest

[b00kiest]

Ok one more thing unrelated to the voices. I don't know if it's just me but the window for the plugin is too large and when I move or try to shrink it, the window disappears and the only way to get it back is to press enter. Like when I use ALT+Tab it does not show up as a window. I have my resolution set to: 2560x1440 so it should be able to handle the window size but it's literally the top to bottom off the edge which is why I try to shrink it by dragging as I can't move it to see the buttons at the very bottom but I can't ever get to them. I have worked a way around it but it is a bit annoying.
Is it something I can fix?

[jackie_w]

Originally Posted by b00kiest

Ok one more thing unrelated to the voices. I don't know if it's just me but the window for the plugin is too large and when I move or try to shrink it, the window disappears and the only way to get it back is to press enter. Like when I use ALT+Tab it does not show up as a window. I have my resolution set to: 2560x1440 so it should be able to handle the window size but it's literally the top to bottom off the edge which is why I try to shrink it by dragging as I can't move it to see the buttons at the very bottom but I can't ever get to them. I have worked a way around it but it is a bit annoying.
Is it something I can fix?

No-one has raised this as an issue before but even if they had I'm not sure I would have been much help. The finer points of Windows configuration is really not my strong point.

In Settings > System > Display > Display resolution my (fairly unremarkable) ASUS laptop says
1920 x 1080 (Recommended)
I don't think I've ever changed it since the day it arrived. As you can see from the attached screenshot my 'TTS to MP3' dialog box is nowhere near fullscreen. The main calibre library window in the background is fullscreen.

In calibre's Preferences > Look&Feel > Main interface tab there is a setting (line 5) labelled
'Adjust for high resolution screens (needs restart)'
I'm not sure what that does but it might be worth investigating.

Originally Posted by b00kiest

Interestinly, I can see the Cortana Voices installed in this folder but they are not listed in the TTS area.

I'm a bit confused. You made 3 posts about MS Voices. Are you still lacking access to the Cortana Voices within calibre or did you solve it?

If the former, there may be other keys in the Windows registry from where you can export the base data. I can dig out my old Cortana notes from when I originally did this work on my own PC.

BTW, Cortana isn't enabled on my PC either but with enough rooting around the registry I did eventually find an entry I could export/edit.

[carlinhOS]

Hi

Today, after a year without messing with Calibre I installed the last version of the program and reinstalled TTS to MP· plugin too.

I am trying to convert many new epubs to mp3, but after job started I get a recurrent erro:

---
TTS to MP3: author - title (747)
Traceback (most recent call last):
File "runpy.py", line 194, in _run_module_as_main
File "runpy.py", line 87, in _run_code
File "site.py", line 82, in <module>
File "site.py", line 77, in main
File "site.py", line 49, in run_entry_point
File "calibre\utils\ipc\worker.py", line 216, in main
File "calibre\utils\ipc\worker.py", line 151, in arbitrary_n
File "calibre_plugins.tts_to_mp3_plugin.jobs", line 95, in do_book_action_worker
TypeError: cannot unpack non-iterable NoneType object

Logfile for 747 / 1 / section-of-xlm-file.xhtml
---

So I tried to convert old epubs I'd already converted without erros a year ago to see what happened, and I got the same error.

Anybody knows what's happening? Is it a Calibre bug or maybe is it related to Windows 10 updates, LAME, regional configuration (my language is spanish)...

Sorry because I can understand python error messages showed in Calibre. I attached my epub file (in spanish), converted succesfully to mp3 a year ago.

[carlinhOS]

Originally Posted by carlinhOS

Hi

Today, after a year without messing with Calibre I installed the last version of the program and reinstalled TTS to MP· plugin too.

I am trying to convert many new epubs to mp3, but after job started I get a recurrent erro:

---
TTS to MP3: author - title (747)
Traceback (most recent call last):
File "runpy.py", line 194, in _run_module_as_main
File "runpy.py", line 87, in _run_code
File "site.py", line 82, in <module>
File "site.py", line 77, in main
File "site.py", line 49, in run_entry_point
File "calibre\utils\ipc\worker.py", line 216, in main
File "calibre\utils\ipc\worker.py", line 151, in arbitrary_n
File "calibre_plugins.tts_to_mp3_plugin.jobs", line 95, in do_book_action_worker
TypeError: cannot unpack non-iterable NoneType object

Logfile for 747 / 1 / section-of-xlm-file.xhtml
---

So I tried to convert old epubs I'd already converted without erros a year ago to see what happened, and I got the same error.

Anybody knows what's happening? Is it a Calibre bug or maybe is it related to Windows 10 updates, LAME, regional configuration (my language is spanish)...

Sorry because I can understand python error messages showed in Calibre. I attached my epub file (in spanish), converted succesfully to mp3 a year ago.

FYI, I saw previous message in this thread https://www.mobileread.com/forums/sh...7&postcount=81 and I am converting from and to Windows Desktop folder, so there are no non-ASCII characters in the path, and the errors persisted.

[carlinhOS]

Originally Posted by carlinhOS

FYI, I saw previous message in this thread https://www.mobileread.com/forums/sh...7&postcount=81 and I am converting from and to Windows Desktop folder, so there are no non-ASCII characters in the path, and the errors persisted.

FYI (2), I'm using the plugin I downloaded from this thread, installed it from file and restarted Calibre. Bug persists.

[carlinhOS]

Originally Posted by carlinhOS

FYI (2), I'm using the plugin I downloaded from this thread, installed it from file and restarted Calibre. Bug persists.

Finally I manged to get my mp3 file.

There must be a bug in the plugin. At least, when any of the metadata fileds of the mp3 file you are creating contains a blank space within, the conversion fails.

For example, I changed 'Album' content from 'L 39-2015 PACAP' (acronym for a spanish law) to 'L39-2015PACAP' and the plugin worked out perfectly.

[jackie_w]

Originally Posted by carlinhOS

Finally I manged to get my mp3 file.

There must be a bug in the plugin. At least, when any of the metadata fileds of the mp3 file you are creating contains a blank space within, the conversion fails.

For example, I changed 'Album' content from 'L 39-2015 PACAP' (acronym for a spanish law) to 'L39-2015PACAP' and the plugin worked out perfectly.

I've downloaded your sample epub and can reproduce the error message. I'll look into it over the next couple of days.

Please can you confirm the following:

• which version of calibre are you using? Both the version number and whether it's 32-bit or 64-bit Windows.
• Are you using TTS to MP3 v0.5.8? It's shown in the window title of the plugin's main GUI screen.

[jackie_w]

@carlinhOS,

The sample you provided doesn't have a cover image. This produced the same error message in 'TTS to MP3 v0.5.8' for me as it does for you.

If I added any old cover image to the calibre library (not to the epub) everything ran without error. Can you do a quick test to see whether this temporarily "fixes" the problem? If it doesn't work then there is more than one problem happening, but I can't replicate it.

Obviously I'll still need to fix this cover issue, whatever the result of your test. Lack of a cover in the calibre library shouldn't cause the plugin to fail.

[carlinhOS]

Thank you very much @jackie_w for your quick response.

After doing and undoing many things I discovered the error was caused by exceeding total length of the path+filename in Windows, not because of your plugin. As I said I installed last version and it runs properly.

Sorry for bothering you and the other members.