Lojban festvox

Intro

Lojban is a logical spoken language with an unambiguous grammar. It can also be transcribed, phonetically, into ASCII characters (the most common way to read it). As there wasn't a text-to-speech for it, I figured I'd give it a go and try to create one. Festival turns out to be a very flexible platform for text-to-speech needs.

Note: This project has been scrapped for another which does it the right way. See below for more details.

The Goods

To make this go, you will need a copy of Festival, the ked_diphone (American male voice, any quality) and lojban_diphone.scm.

Complete copy of everything I could find related to this topic
festvox_all.tar.gz

Using

Run Festival and type:

(load "/path/to/lojban_diphone.scm")
(voice_lojban_diphone)
(SayText "coi rodo")

Example

You can hear some sample text of what it sounds like. This is a recording of xirli'u selsanga: xirli'u selsanga.ogg in oggvorbis format.

Others

This is most certainly not the correct way to do things. Ideally, there should be a set of diphones for Lojban specifically. Apparently there has been work on it, but I don't know how useful it is.

I've recently been informed that someone has recorded diphones for Lojban. The problem is that there's a little bit of work left: mapping the diphones out. It's been partially done, but the timings are still missing. Hopefully someone can go through and clean it up, but that's a long and tedious process. It's getting done, though. See above.


Steve Pomeroy
Last modified: Thu Feb 27 10:41:07 EST 2003