Re: Making a Vintage Sounding TTS Voice
@musicalman Thanks for the thorough response! I have definitely been playing around with the idea of singing. In fact, singing is much easier to do than natural speech as it is much more confined. I added pitch the other day so you can supply a list of target points across the utterance. Unlike DECTalk, you specify the list of F0 targets independently from the list of phones, so you could make a vowel that is stretched out but which changes notes all over the place. I will prepare something to show this over the weekend.
The Voder was amazing. In fact, it is the first machine that is demonstrated in this rather interesting vinyl recording from 1986 by Dennis Klatt, the author of DECTalk:
https://www.dropbox.com/s/ky5vxfnpc79a9 … ng.au?dl=1
Speaking of vocoders, I did write one of those as well a few weeks ago. Haven't decided what to do with it yet but it sounds pretty cool. You can do the traditional singing synthesizer effect, and if you use a generated carrier you can also do time stretching where the speech gets slower or faster. I might release it as an open source library or a VST plugin at some point, not sure yet.
Regarding unvoiced sounds and transitions, both of these things are definitely work in progress. I did make some improvements to both last weekend but didn't have time to write another diary entry; look out for one this weekend.
@x02 Both DECTalk and ESpeak are formant synthesizers. Klatt synthesis is a style of formant synthesis where you have a bunch of defined parameters that you can play with to make speech sounds. It was used in the MITalk system from MIT, it was used in DECTalk and possibly in a few others - I'm not entirely sure. Both ESpeak and DECTalk are formant synthesizers, in other words, just based on different techniques for how to generate the excitation signal and a few other details as far as I know.
I do actually have a public domain implementation of a Klatt synthesizer, but I didn't like the sound of it so I wanted to make one from scratch using the techniques I knew. Whether I will get all the way remains to be seen, of course, but it's a lot of fun to have it as a side project regardless of how it turns out.
There is indeed a repository but I haven't made it public. I just didn't edit the mentions of it out of the diary, as I wanted to keep it intact. I haven't decided what to do with the implementation yet; it really depends on what kind of quality level I am able to reach.
Kind regards,
Philip Bennefall
-- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
