Luciano Esteban Notarfrancesco wrote:
I have roughly implemented -enterely in Squeak- a synthesizer based upon rsynth (Klatt cascade/parallel filter bank synthesizer and rule-based text to phonemes translator), but it's just an > experiment...
Wow!
Rsynth is public domain I believe. Do you have any plans to put your work with it under the Squeak license (or something similar)? If so, I'd be interested in helping out as a tester or with comments related to getting it to work efficiently or better under Squeak.
We have a Delphi product (an interactive story authoring system called StoryHarp) which I'd like to see in Squeak (we almost wrote it in Squeak in the first place). It uses speech recognition and text-to-speech (although the speech recognition is not as essential).
A native Squeak test-to-speech (TTS) system would be much easier to manage across multiple platforms than an API w/ primitives to call out to platform dependent TTS systems.
the quality is pour and it's very slow. I'd like to do something in the diphone concatenation direction (just like Festival does, for instance).
How slow is it? What system (processor/speed/memory) are you running it on? Is it as understandable as the original rsynth itself? Are there specific Squeak speed related problems (like clicking or drop outs or stuttering) that could be worked around?
Is it possible to use the Smalltalk->C translator to take your work and build it into the VM (like some current sound primitives)? It would have to be written with the translator restrictions in mind, or the translator would have to be expanded to support it.
Festival is only free for non-commercial use I believe. If you used that code directly, your work could not be part of the general Squeak distribution under the Squeak license (although I guess it could be distributed as an add on).
I'd like to know of other people working in the same thing for Squeak, if there are any.
I'd be very interested in finding this out too.
I've looked at linking Squeak to SAPI and Microsoft Agent under Windows via Squeak primitives -- but that would just provides access to external TTS and SR systems. I did not get anything working though because I decided to use Delphi instead. I don't know much about implementing TTS or SR engines themselves, but I've done some signal processing stuff for machine vision and acoustic positioning.
It may sound silly, but TTS is a real whiz-bang feature that would attract more attention to Squeak. People are very excited by Macintalk on the Newton for example. Also, I've heard one can also use a good TTS system to generate wave forms to use as the basis of a speech recognition algorithm. Your work could enable a lot of really neat stuff.
-Paul Fernhout Kurtz-Fernhout Software ========================================================= Developers of custom software and educational simulations Creators of the Garden with Insight(TM) garden simulator http://www.kurtz-fernhout.com