KSpeak - eSpeak TTS porting for kindle

[notimp]

Originally Posted by encol

KOReader is not a kindlet. Kindlets' permissions have to be elevated to access to the fs.


and where you attached the sound card?

USB sound card dongly thing. (To the only port on the device.. )

(edit: Missed the bold part - > Yeah, thats the thing... sshing into the Kindle using USB wasn't such a bright idea, when the command you are testing produces a sound output, that has to be checked with the usb sound card dongle thats - at that point not connected to the device.

So I removed *usb cable* and attached *soundcard dongle* and input the commands in kterm instead. Genius. And yes - that worked. Heard the TTS sound output, could use it with different languages... - So the issue is related to privilege escalation on the kindlet itselft. )

I'll wait for additional feedback. If no one else is having problems getting the .azw2 to work - I'll downgrade>jailbreak>upgrade my Kindle again - maybe...

For now - using the shell command I could listen to the german TTS voice - and, lets just say, its no Ivona.. (but then the install package was only a few MB, so I saw that one coming.. ).

But then - its miles better than Amazons own approach of *beeping* into your ear with intensity on every page change...

[notimp]

Tried the KSpeak on a Kindle Touch I had lying around, just to see if the language files would resolve - same issue.

I then re-installed kindle-mkk-20141129 (appropriate for that kindle touch) - same issue.

Likelyhood of both Kindles suffering from the same strange mkk permission escalation error. Slim.

[knc1]

Check what user you are, if 'framework', then you need to switch user (su) to 'root'.
One of the things different about the newer 'hotfix' is that it installs a patched version of su that does not require a password.
(In addition to the stuff that was contained in MKK)

KUAL is a worked/working example of how/when to test for and conditionally switch user to 'root' in its start-up scripting.

That is:
The switch user isn't automated, if you need it, you will have to call it. We just provide it.

[notimp]

whoami via kterm on both the PW3 and the KT comes up as root
(edit: but then, kterm is launched via kual)

[knc1]

Originally Posted by notimp

whoami via kterm on both the PW3 and the KT comes up as root
(edit: but then, kterm is launched via kual)

That is good - indicates that everything you need is included and working.

Now, on the new *.azw2 -
(Which isn't launched by anything except a touch on it.)

Set it up to speak the result of whoami.
run it.
see what it has to say about itself.

See the example of using flite to speak the output of commands that is included with the flite install.

Note:
Vague with embedded flite was built in the old days, before we needed to do a su to root.
So it will work from KUAL now (which does the switch) but it may not work from the new Kindlet as framework.

[notimp]

whoami | /mnt/us/extensions/espeak/espeak.sh

works from kterm (gives the voice output "root") - but I'm afraid packaging that as a kindlet (.azw2) is above my skill level.

(I could learn, but when the initial barrier to entry is "not enough diskspace to install XCode on this Mac" I guess, I might resort to not setting up a dev environment... )

At least I've given you some tester feedback, if it turns out that more people are having this issue.

[knc1]

Am I really lost now?

I thought the trouble report was that it did not work with the *.azw2 that was made to control it?

So we seem to have eliminated all other possibilities, other than the new application document not doing the switch user call.
I can't think of how to work around that, the document will have to be fixed.

[notimp]

I'm not the author - and I'm only "one user" so far, reporting, that the .azw2 doesn't work for me.

The .azw2 basically acts as a gui front end to send along a few parameters (like language, text, or filepath) to a shell script (as far as I understand).

And yes, the azw2 file crashes with an error massage, as soon as I open it. Even exiting it, the Kindle framework will write an "app could not be exited cleanly" error message.

The trouble shooting process was to try to eliminate potential causes - I've tried two different Kindles, with two different (re-) installations of MKK, both of witch run other kindlets (KUAL, JBPatch, ..) fine. Also - everything from the script level downstream seems to work as well -

but the issue might be hard to replicate for the author, or still be an isolated problem - although its less likely to happen on two Kindles.

Thats basically it - I hope I helped with the error assessment, but I can't "fix it". And the original author might not either, if he is unable to replicate it. User account switching not happening might be the cause - but I can not say that with certainty.

[knc1]

Not lost, that was what I understood also.

I might have made a poor choice of words, making it sound like something other than suggestions for whoever (the author) might be interested.

= = = = =

Making it a feature of KOreader (which we have the complete source and control of) would probably be easier than trying to add it to the native reader.

I post that in response to the question about extracting text from books and/or documents.

That is one of the basic functions of a 'reader' program, so why re-invent that wheel?
(the other basic function is to display what it has extracted)

[encol]

Anyone else has tested the kindlet?

Originally Posted by knc1

Making it a feature of KOreader (which we have the complete source and control of) would probably be easier than trying to add it to the native reader.

This is a good suggestion, but i prefer the native reader and i don't feel like learning lua at the moment

[PoP]

http://www.festvox.org/flite/

Originally Posted by encol

Anyone else has tested the kindlet?
...

Yep, on PW3 5.8.7. For me,

the .sh scripts work fine under kterm.

the .azw2 frontend doesn't crash or give error message, but I can't hear any tts from it.

[EDIT] Nice port. espeak seems to use the mbrola voices and seems to transliterate text to mbrola phonemes. that is good for tts in many languages (which flite doesn't support). Using mbrola natively requires the input to be phoneme based and is not practical for tts but with effort it can even sing!.

[encol]

Originally Posted by PoP

the .azw2 frontend doesn't crash or give error message, but I can't hear any tts from it.

try to speak something from the kindlet and post KSpeak.log, please

[PoP]

^

Empty: developer/Kspeak/work/Kspeak.log

KSpeak.log.zip

Tried twice. Maybe I don't know how to trigger tts? Is "return" okay after entering "Hello World" into the text box?

[encol]

You have to click the mouth icon
Tap the title bar (or something else) to hide the keyboard

[notimp]

I updated the PW3 to 5.8.7, now the azw2 file doesnt crash. So it was either a version dependency, or the update fixed some unknown issues that manifested themselves only in KSpeak not working..

KSpeak is very basic at the moment.

- Opening a text document located in a subfolder of /documents - that contained spaces and a hyphen in its filename didn't work (might also have been pathlength related).
- Canceling a TTS session is not yet implemented, when starting another one - the audio outputs just get layered over each other.
- (Re-?)Connecting the USB soundcard dongle while KSpeak is open results in the Kindle message popping up that tells you to set the Kindle to english, if you want to use TTS. Clicking that away (or clicking on it while a TTS session is still running?) results in the Kindle softlocking (KSpeak crashes). Turning the display off and on again - will get you out of the softlock (Kindle cant load KSpeak and exits you back to the menu) - but after about five seconds, the kindle will announce a reboot (because it noticed KSpeak crashing), and autoreboots if you don't cancel that process.

Rebooting with the audio dongle plugged in causes the Kindle to launch recovery (because of the usual mobileread recovery scripts), getting out of recovery is a hassle for people that don't know how (first exit listed on screen does nothing, normal exit bootloops, you have to find the right "exit and disable recovery" option) -

people without eyesight are locked out of their Kindles at that point, unplugging the USB device will not fix it (maybe it will in connection with a cold reboot, but afair, the reason the recovery is launching/boot looping is because a textfile gets created in user root, and that doesnt vanish), ...

There is no way to follow reading/text progression on screen (I didn't have the expectation, but just for people to know before purchasing a dongle), there is no way to selectively start the TTS at a certain point in a longer text, there is no way to stop TTS, once initialized, other then unplugging the dongle (and at that point the process might still run in the background - so at least I chose to reboot afterwards), ...

The german TTS is basically unusable for the purpose of having it read out long texts to you, ...
--

So thats KTerm from a user experience perspective.

Its still a neat thing to have - and it's great that it has been realized on the device - but from a usability perspective - don't get your hopes up.

I you NEED TTS to read long form texts to you - use an Android device, get the IVONA (now owned by Amazon) voices (free, but some of them have been depublished from the Play Store since the corporate takeover, but you can still get them at apkpure for example (you have to rename the container to .zip, unzip it and move the files to their proper locations in afair /Android/obb/)) - then use a program like FBReader (I'm using version 2.5.10 (because newer ones are "too dumbed down" for my purposes), and its TTS plugin to use Ivona voices to read epubs (that can come right out of the Android Dropbox app, but on't have to.. ) -

Its free to set up, gives you the best TTS quality out there (imho, but as an educated statement, also Amy (UK english) is far better than the US english pendent - https://www.youtube.com/watch?v=jRUb_wlILBg ), once you tweak reading speed a little (possible in FBReaders TTS plugin) to make it faster. Saves reading progress, allows bookmarking - auto scrolls the text in parallel, and allows you to invert text colors, if you'd wan't to do that.

Also - as mentioned before, Amazons own Kindle PW3 TTS engine (english only) is unusable, because they decided to *peep* you on every page change. And that turns out to be EXTREMELY annoying after lets say page 3 or so....

So long story short - if you NEED good TTS functionality - buy an Android device.