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

[jackie_w]

@KindleAZWorKFX,

I'm not sure what else I can say that I haven't already said. I don't have the technical skill-set to write a plugin which will interface with Chrome or any other web-based app.

@BetterRed,

As for Balabolka, I looked at the main website yesterday. I didn't see any teasers about plans to work with cloud-based voices - but that doesn't prove anything, of course. Let's wait and see.

Perhaps a relevant point is that I do have this app loaded on my PC (as a portable app rather than fully installed), but the only voices available by default were the ones in the

[HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\Speech Server\v11.0

section of the Windows registry, which are pretty old, and pre-date the Win10 voices. By doing a bit of (non-user-friendly) registry hacking (similar to those referenced in post #2) I was able to access my purchased Ivona voices (the 32-bit versions of them) but I didn't find any Balabolka option to make scanning for these extra voices easy. Did I miss something? On my 64-bit PC, using 64-bit calibre, this plugin as-is locates the Ivona voices (the 64-bit versions of them) without having to do anything at all.

On a more general point the handling of voices in Windows looks like a bit of a dark art. There is endless repetition of voice code throughout the registry. Even in the MS Edge section the same voices are listed in two different places. This is the key for one of them on my PC

[HKEY_CURRENT_USER\Software\Classes\Local Settings\Software\Microsoft\Windows\CurrentVersion\AppContainer\Storage\microsoft.microsoftedge_8wekyb3d8bbwe\Children\001\Software\Microsoft\Speech_OneCore\Isolated\k6BwW5fty3XWMlmQDjzUq77TXIsLVjJGt268p8btklc\HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech_OneCore\Voices\Tokens\MSTTS_V110_enGB_SusanM]

Delightful! If, when they arrive, the new cloud-based voices turn up in the same place then I'll see them and do some tests but the "cloud-based" description makes me wonder whether there will be any local files installed at all.

[jackie_w]

• Maintenance upgrade for python3 compatibility.
• Plugin icon updated to be more visible when using calibre's Dark mode.

Plugin should now work with both calibre current production version (v4.15) and calibre python3 beta versions v4.99.x.

If any problems, please report them here.

[Fiat_Lux]

Any chance of this being modified to work on Macs or Linux boxes?

[jackie_w]

Originally Posted by Fiat_Lux

Any chance of this being modified to work on Macs or Linux boxes?

Not by me, I'm afraid. The plugin's existing GUI functions could probably be re-used by Mac/Linux without too much effort but it's the OS text-to-speech components which are the showstoppers.

There may be Mac/Linux utilities around which could replace the LAME.exe (free) Windows utility which converts from .WAV to .MP3, but the utilities necessary to do the initial conversion from text to .WAV (including the Voices used) are part of Windows itself.

If there is similar built-in functionality and Voices for Mac and/or Linux it would need someone who is familiar with those OS's to do the research and work.

[DictatorAlly]

Linux user's could use something like this:
https://github.com/mozilla/TTS

Ideally you'd run a docker with an imported model already trained and then have the calibre plugin interface with that.

Google's TTS is rather expensive at $16/1m characters, the first 1m/month is free though, perhaps an implementation to that would be worthwhile as well.

[nickmik123]

is it possible to create subtitles (I'm using VLC to play the mp3's). What output format should be used?

[jackie_w]

Originally Posted by nickmik123

is it possible to create subtitles (I'm using VLC to play the mp3's). What output format should be used?

Aren't subtitles something you add to video files rather than MP3 files?

As I've already outlined in post #12, this plugin is just a GUI wrapper to:

• Use MS Windows SAPI5 Text-to-Speech builtin utilities to create a .WAV file from a plain text string extracted from a single chapter of a calibre book.
• Use the commandline version of the free LAME.exe Windows utility to convert .WAV to .MP3 with appropriate MP3tags for the book.

I don't know of any LAME options which would create a subtitles file.

I seem to remember that I did consider adding the chapter text as MP3 lyrics, but, at the time, I had no success being able to do this automatically during MP3 creation. I'd be willing to look at that again if anyone wants to share knowledge about how to do it with commandline LAME (if it's possible).

[kovidgoyal]

@jackiew: the next release of calibre is no longer bundling pywin32 so you can tuse win32com/win32process/etc anymore. If there are particular APIs you need wrapped, i can add them to the calibrewinutil module, if you send me a list.

[jackie_w]

Originally Posted by kovidgoyal

@jackiew: the next release of calibre is no longer bundling pywin32 so you can tuse win32com/win32process/etc anymore. If there are particular APIs you need wrapped, i can add them to the calibrewinutil module, if you send me a list.

I think I understood the gist of your post to mean that this plugin will stop working soon unless I do 'something'. And that you're offering to help by providing some additional 'somethings' in calibrewinutil.

But, I'm afraid I have no idea exactly what you're asking for in the bolded bit. Can you ask it in layman's terms, please?

Are you asking which win32xxx modules I'm actually importing from and which specific parts of them I actually reference in the code? Or are you asking for something else entirely?

The problem I have is that most of this plugin's low-level win32 code was inherited 'as-is' from someone else's calibre Viewer TTS plugin. I presume they understood it. Unfortunately they're long gone and my understanding of the win32 stuff is minimal. I realise this isn't ideal ...

[kovidgoyal]

Originally Posted by jackie_w

Are you asking which win32xxx modules I'm actually importing from and which specific parts of them I actually reference in the code? Or are you asking for something else entirely?

Yes, exactly. I can wrap those same things inside the calibre winutil module so you can use them from there, instead.

The problem I have is that most of this plugin's low-level win32 code was inherited 'as-is' from someone else's calibre Viewer TTS plugin. I presume they understood it. Unfortunately they're long gone and my understanding of the win32 stuff is minimal. I realise this isn't ideal ...

Hmm, well that is unfortunate.

[jackie_w]

Originally Posted by kovidgoyal

Yes, exactly. I can wrap those same things inside the calibre winutil module so you can use them from there, instead.

Hmm, well that is unfortunate.

I've tidied up the existing win32 import statements and checked that the plugin still appears to be functioning OK on Win10 after the changes:
- calibre 64-bit v5.2
- calibre 32-bit v4.23.

I believe this is the full list of win32 imports:

from win32com.client import WithEvents
from win32com.client import Dispatch
from win32com.client import DispatchBaseClass
from win32com.client import CoClassBaseClass
from win32com.client import CLSIDToClass
from win32com.client import util as w32comclient_util

from win32com.server import util as w32comserver_util
from win32com.server.policy import EventHandlerPolicy

from win32process import CreateProcess
from win32process import STARTUPINFO

from win32event import WaitForSingleObject
from win32event import INFINITE

from win32con import STARTF_USESHOWWINDOW, SW_HIDE, CREATE_DEFAULT_ERROR_MODE

In case it's relevant, these 2 imports are also used in the '5000-line black box'. They're not ones I've ever knowingly used elsewhere in my other plugins.

import pythoncom
from pywintypes import IID

[kovidgoyal]

Hmm that's a rather large list. Will look into it, can you post your cleaned up plugin.

[jackie_w]

Attachment 182895

Originally Posted by kovidgoyal

Hmm that's a rather large list. Will look into it, can you post your cleaned up plugin.

See attached zip.

The 'big black box' is the tts_typelib.py file. The other 2 files referencing win32xxx are utils.py and other_dlgs.py.

I believe its possible that

from win32process import CreateProcess
from win32process import STARTUPINFO

can now use subprocess instead now we're at python3? CreateProcess was only ever used as an alternative because some of the commandline mp3-tagging args for lame.exe (e.g. Author and Title) could have non-ascii values. I'm looking at this at the moment.

Also, the last 2 lines of the original list are all integer constants so I suppose I could just hard-code their values??? I don't know if they are subject to change over time.

[jackie_w]

@Kovid,

Is it safe to assume that for users of calibre version <= 4.23, the win32 modules will always be available?

If so, I believe I now have subprocess working OK in python3/calibrev5.2 with unicode args. In which case the last 5 rows can be removed from my original list

from win32com.client import WithEvents                                  #function
from win32com.client import Dispatch                                    #function
from win32com.client import DispatchBaseClass                           #class
from win32com.client import CoClassBaseClass                            #class
from win32com.client.CLSIDToClass import RegisterCLSIDsFromDict         #function
from win32com.client.util import Iterator as w32comclient_util_iterator #class

from win32com.server.util import wrap as w32comserver_util_wrap         #function
from win32com.server.policy import EventHandlerPolicy                   #class

from win32process import CreateProcess
from win32process import STARTUPINFO

from win32event import WaitForSingleObject
from win32event import INFINITE

from win32con import STARTF_USESHOWWINDOW, SW_HIDE, CREATE_DEFAULT_ERROR_MODE

ETA: All imports are now either a class or a function (rather than its module).
The attached updated plugin still seems to work OK in v4.23 and v5.2

[kovidgoyal]

Yes, for 4.x they are. I will take a look at the rest.