Text To Speech Support about macchina

[uttarayan21]

Sure no problem ! Good luck !

[grtcdr]

Your bindings work! 🚀 😃

[uttarayan21]

Maybe I should update the dependencies and properly add directions

[uttarayan21]

Then I guess I'll publish it in crates.io
I just published it in crates.io google_speech

[uttarayan21]

Well I have a library or more specifically rust bindings to a python library. https://github.com/uttarayan21/google_speech_rs that makes it super easy to do tts. But I haven't tested it super extensively.

Also you swapped the () [] around in the link lol

[grtcdr]

Also you swapped the () [] around in the link lol

😆

Well I have a library or more specifically rust bindings to a python library. https://github.com/uttarayan21/google_speech_rs that makes it super easy to do tts. But I haven't tested it super extensively.

I'll be allocating a little bit of time to try and get this to work, and see if we can have this feature out with the next release.

[grtcdr]

Is it on crates.io?

[uttarayan21]

No I didn't really publish it since I since I didn't test it extensively.

[grtcdr]

I'll just clone and test it locally then, thanks for sharing! 🚀

[grtcdr]

Any dependencies I should be installing? I'm getting a ton of build errors? Ignore this.

[grtcdr]

Google's guide mentions that one should install sox and libsox-fmt-mp3, installing sox by itself and running your example results in:

sox FAIL formats: no handler for given file type `mp3'
PyErr { type: <class 'RuntimeError'>, value: RuntimeError(), traceback: Some(<traceback object at 0x7f00cbe21440>) }

libsox-fmt-mp3 is not available in the official Arch repos or the AUR.

A quick sudo find / -name "*sox*" finds:

...
/usr/lib/sox/libsox_fmt_mp3.so
...

So Arch does package the libraries required by google's speech python library, but unfortunately it doesn't work.

[uttarayan21]

Optional Deps : libao: for ao plugin
libmad: for mp3 plugin [installed]
libid3tag: for mp3 plugin [installed]
wavpack: for wavpack plugin
libpulse: for pulse plugin [installed]
opusfile: for opus plugin
twolame: for mp3 plugin [installed]

They are listed as optional dependencies
libmad twolame libid3tag

[grtcdr]

I missed these, sorry!

[grtcdr]

I'm all for you publishing this on crates.io, it works + it's great!

[grtcdr]

Awesome! I'll starting using that right away then :)

Thanks 👍

[grtcdr]

Okay, so I managed to make it work, but I feel like it's a very hacky method...

[grtcdr]

It takes into account the configuration of course, but ignores bars as I don't see a way we can make google_speech speak "bars" lol.

A good improvement is to understand which readout will be read and do some processing on the data to make the speech a lot more understable.

For example, instead of reading each digit individually in "5.10.46-1-lts" it should read it like any normal human being lol.

We can also make it so it says "Kernel is Linux 5.10.46-1-lts" instead of "Kernel" pause "Linux 5.10.46-1-lts"

[grtcdr]

Here we go #110.

It doesn't actually ignore bars, but I'll fix this right away, becuase apparently GoogleSpeech can speak in bars, and it goes something like this:

Filled circle, filled circle, filled circle, filled circle, filled circle...

Possibly the most hilarious thing I've ever come across developing macchina.

[uttarayan21]

Filled circle, filled circle, filled circle, filled circle, filled circle...

Haha I never knew lol ! That is hilarious.

For example, instead of reading each digit individually in "5.10.46-1-lts" it should read it like any normal human being lol.

I think that is due to it thinking that the point's are decimal places, however no one has any control over it since the text is send to google servers ant that is what they reply with.

Also about espeak and other similar offline tts engines don't sound quite as natural so that is the reason I made this.

[grtcdr]

I think that is due to it thinking that the point's are decimal places, however no one has any control over it since the text is send to google servers ant that is what they reply with.

Ah okay!

Also about espeak and other similar offline tts engines don't sound quite as natural so that is the reason I made this.

I don't know too much about other TTS engines, and although some won't agree with the fact that we're using Google, if it's the most natural sounding engine, then I'm fine with having it in.

Plus, I think GoogleSpeech works offline? I tested it offline and it still reports data without an internet connection.

[uttarayan21]

Plus, I think GoogleSpeech works offline? I tested it offline and it still reports data without an internet connection.

If it has same data as a previous call then it is cached. If you try with text that hasn't been said before it will fail if offline.

[grtcdr]

I thought about that, thanks for confirming it.

[grtcdr]

Closing this as #110 was merged a while ago.