Choose or invent a container format that has less overhead than the one
you currently use, and that contains only elements necessary in your
application.
Or, store multiple words (or maybe even the whole corpus) in the same
Speex file, and then seek to the right position before playing back, and
only play back as much as you need to (the word).
You aren't trying to make a speech synthesizer by any chance though?
Anders Johansen wrote:> Hi,
>
> I am using a digitized voice with a vocabulary of some 350.000 words.
> Converting them from WAV to Speex obviously cut down the size a lot,
> but I find that some more compression could be done. Specifically I
> collect the resulting files in one big container file, and that one
> compresses quite well (from about 600MB to about 130MB). I expect that
> is due to redundant data in the headers and other OGG container
> artifacts (framing etc.).
>
> Any suggestions on how to proceed?
>
> Sincerely,
> Anders S. Johansen
> _______________________________________________
> Speex-dev mailing list
> Speex-dev@xiph.org
> http://lists.xiph.org/mailman/listinfo/speex-dev