Using non-speex files for voice

[jenova314]

Hi all,

I recently upgraded my H320's firmware to 3.x, and am disappointed that I can no longer use MP3s as my .talk files.  Is there any way around it, aside from downgrading?  Many of my directories are either in Mandarin or Japanese, and speex does an extremely horrible job at pronouncing them.  I imagine the switch to speex was to save space, but at a great cost in voice quality IMHO.

I used to use these types of sites to generate my voice files in different languages, and have been pleased:

http://text-to-speech.imtranslator.net/

[Llorean]

"Speex" is not a text-to-speech Engine. It doesn't affect the pronunciation, all it is is a different way of making the .wav files you get smaller (MP3 is not as effective for voice).

[jenova314]

Well, that explains a lot of the trouble I've been having!  Thanks!

[bluebrother]

Btw, I suggest checking first -- there is a nice explanation of speex at wikipedia: http://en.wikipedia.org/wiki/Speex

[jenova314]

Thanks for everybody's help.

I've compiled some wav files that I would like to encode into speex files.  However, having used the Speex ACM Codec 1.0.1.1, dBpower Speex Codec, or Speex Frontend UI, I have yet to yield a single file that was able to have my H320 say the file directories.  I have renamed the files appropriately to _dirname.talk under directories that I'd like voiced, and it is simply not working.

However, when I use the rockbox util, it works just fine.  Strangely, the files generated by rbutil does not play in foobar2000, whereas my self-encoded spx files do.  Unfortunately, I do not like the TTS included with rbutil, and would like to use my own.

Am I doing something wrong?  Also, does anybody know how I can have rbutil use different TTS in different languages?  Thanks!

[nls]

This is not strange at all since rockbox uses a raw speex format while regular spx files use a container (Ogg), so rockbox talk clips and regular speex files are incompatible. I'm not sure how it works but rbutil should be able to interface with sapi voices to create talk clips.

[gevaerts]

If you have a rockbox build environment set up, you can build rbspeexenc, which is a commandline tool that encodes the files correctly for rockbox. To do this, go to the tools/rbspeex directory, and run make. rbspeexenc will then be built in tools/

If you can't compile programs yourself, please let us know, and tell us which OS you're using. Somebody should be able to compile for you then

[jenova314]

Sorry, I'm not very savvy, and I wouldn't trust myself to complile anything.

I'm running Windows XP with SP2, 32-bits.  Does that help?

[bluebrother]

You can use any SAPI voice with Rockbox Utility ...

[gevaerts]

I've put a compiled version for windows at http://files.hostname.be/rbspeexenc.exe

You need to run this inside a command prompt (cmd.exe). It will print out usage instructions if you run it without arguments.

However, make sure you really can't get rbutil to do what you want first, as using rbutil will very probably be a lot easier.

[jenova314]

Hey gevaerts,

That worked perfectly, thanks!  Now if I could just find the SAPI in a form that rbutil will accept...

[bluebrother]

That worked perfectly, thanks!  Now if I could just find the SAPI in a form that rbutil will accept...

SAPI is a programming interface that is used by rbutil (well, the vbs it executes actually). There is no special SAPI -- either your voice is SAPI compliant or not.