Google Translator and accented characters problem

[anhakkar]

First of all, what a great plugin this is!

What is really cool that we can use Google Translator TTS engine to make it talk in any language. But there is a problem. If I use accented characters, like ä, in a phrase, those characters are encoded in a way that doesn't make sense to Google. For instance, if you want Homeseer say, "Motörhead", it sounds right with Homeseer voice, but not with Google translator voice. Could this be fixed?

[alexbk66]

I'll have a look, sure.

And I love Motörhead, my favourite song God Was Never on Your Side

[alexbk66]

I made the Google TTS text encoding configurable, so accented characters are supported (not encoded). But this change may produce unexpected results, i.e. '+' sign is not encoded anymore, so it's now ignored because it's treated as 'space'. URL encoding is a messy question, so I don't know which way is safer. We'll see..

Ver 1.0.0.7 now in Beta section (until approved by HST)

[XeKToReX]

Is it possible to get localised voices? Like English (AU) for us Aussies, Alex?

Guessing not because Google translate TTS isn't Google Assistant's voice.

[alexbk66]

Is it possible to get localised voices? Like English (AU) for us Aussies, Alex?

Actually it's pretty easy, you can test yourself using Google Translate url, i.e.

https://translate.google.com/transla...guage&tl=en_uk
https://translate.google.com/transla...guage&tl=en_us
https://translate.google.com/transla...guage&tl=en_au

I'll see when and how I can modify the plugin to support that.
Also I'm planning to add official Google Cloud translate API as another option - it should be much better, but will require user registration.

Anyways, better place for requests:

https://forums.homeseer.com/forum/me...atures-request

[alexbk66]

Ver 1.0.0.8 - Couldn't resist and implemented "en_us", "en_uk" and "en_au", and "CUSTOM" setting for any other languages, where you can enter the language id string directly.

[anhakkar]

I made the Google TTS text encoding configurable, so accented characters are supported (not encoded). But this change may produce unexpected results, i.e. '+' sign is not encoded anymore, so it's now ignored because it's treated as 'space'. URL encoding is a messy question, so I don't know which way is safer. We'll see..

Ver 1.0.0.7 now in Beta section (until approved by HST)

This is an old thread, I know, but… Did you really implement this? I may be missing something but I fail to see this configuration option anywhere.

[alexbk66]

This is an old thread, I know, but… Did you really implement this? I may be missing something but I fail to see this configuration option anywhere.

I did make it configurable at some stage, but then found better solution (can't remember exactly).

Doesn't it work?

[alexbk66]

This is an old thread, I know, but… Did you really implement this? I may be missing something but I fail to see this configuration option anywhere.

Actually, I checked the 1.0.0.7 commit - I added the setting in the config file - not exposed to the user because it was quite untested.

If you wanna try, it's in HSPI_AKGoogleCast.exe.config, change <value> to True

<setting name="UrlEncode" serializeAs="String">
    <value>False</value>
</setting>

[anhakkar]

I tried this, but it doesn't seem to make any change in the output. I have a test event in HS: Speak "Motörhead". Either way, the Googlecast speaker says something like this:

mot, percent, U, 0, 0, F, 6, R, H, E, A, D

That is, mot%u00F6rhead

It seems that somewhere, URL escape codes are not formed correctly. Escape codes for urls and html are not the same.

https://graphemica.com/ö


This url, when pasted to browser, sounds right:
https://translate.google.com/transla...rhead&tl=fi_fi


This is what it now sounds like:
https://translate.google.com/transla...rhead&tl=fi_fi

[alexbk66]

I tried this, but it doesn't seem to make any change in the output. I have a test event in HS: Speak "Motörhead". Either way, the Googlecast speaker says something like this:

mot, percent, U, 0, 0, F, 6, R, H, E, A, D

That is, mot%u00F6rhead

It seems that somewhere, URL escape codes are not formed correctly. Escape codes for urls and html are not the same.

https://graphemica.com/ö


This url, when pasted to browser, sounds right:
https://translate.google.com/transla...rhead&tl=fi_fi


This is what it now sounds like:
https://translate.google.com/transla...rhead&tl=fi_fi

"Motörhead" - that exactly what I was using for testing

I'll have to check again... I remember that .NET provides many different functions for encoding url, all of them have different behaviour...

[alexbk66]

if (UrlEncode)
    text = HttpUtility.UrlEncode(text, Encoding.Default);
else
    text = text.Replace(" ", "+");

I'm using HttpUtility.UrlEncode

Did you restart the plugin after changing "UrlEncode"?

[anhakkar]

I did. In fact, I had to shutdown HS to be able to edit the file.

[alexbk66]

I did. In fact, I had to shutdown HS to be able to edit the file.

Ok, I'll check again.