« previous post | next post »

A cool demo video from the MAGE pHTS project:

MAGE/pHTS, a real-time speech synthesis library from NUMEDIART on Vimeo.

The announcement came from Nicolas d'Alessandro and Thierry Dutoit.

Both MAGE pHTS and the underlying HTS synthesis system are open source. If you aren't familiar with development in a Unix environment, putting together your own version may be a bit of a learning experience, especially since you'll need to install HTS as well, choose or train voices, add a text-analysis engine (e.g. from Festival) or create your own input files, etc.

But learning is good, right? And look at how cool the results can be.

Extra points to whoever first connects it up to be controlled by:

1) a theremin;
2) an xbox Kinect;
3) (some signal processing applied to) EEG.


  1. kktkkr said,

    October 5, 2011 @ 9:32 pm

    Reminds me of the voice distortion used in the Portal series of games. These things always crack me up. Could make some brilliant April Fools' jokes. It could be the next auto-tune.
    I suggest hooking this up to real speech samples or perform minor adjustments on the phoneme sequence. Also, this might be a path to better robot singing.

  2. Chh said,

    October 5, 2011 @ 10:01 pm

    Looks like it's controlled in Max/MSP. I'd love to see how they put those together.

    I took some time last year to get Chris Darwin's sine wave speech script for Praat to generate tables of data readable in Max, and now I have a little program that makes melodies out of formant contours…

    It's cool to see a few nice pieces of technology put together in this!

RSS feed for comments on this post