syng3 beta team

[xoxos]

syng3 is currently in alpha and likely to reach beta in the weeks ahead.

looking for folks with a preexistent extensive body of work with a voice synth, eg. syng2 or vocaloid and the advanced sensibility for concatenation and expression. please post with links, and probably i'll pick someone else because i'll only make a few special interest sales and want to pony as many of you as possible, right

my weasel brain also deduces that i could use this process to connive demo audio out of suckers for free. volunteer now, for america!

i will say succinctly that i have experienced a somewhat rarified appreciation, like when you manage to paint something good with an operable brush, like hey this brush didn't suck when i painted with it. or it's a good medium that can be sculpted to an appreciable precision.

[spacedad]

i'd like to give it a try,but i don't have a lot of spare time atm.(win7/eXT)

[seedpress]

xoxos

I've been working on a Faust http://faust.grame.fr/ coded rendition of a Kelly Lockbaum, source-filter waveguide. (Just found your .sem version, yesterday.) My own model is derived from that of Leonard Manzara, of the GnuSpeech http://www.gnu.org/software/gnuspeech/ project. My overall goal is considered batty by most folks, but here it is: The Dolittle Project http://www.dolittleproject.com

I'm not looking for any contract work. I'm basically a hardware guy. That's why Faust was appealing: coding is much like drawing circuit diagrams. You are obviously in the Einstein class of programming skills, whereas I'm in (maybe) first grade. On the other hand, I believe the overarching goals of the The Dolittle Project have a better chance of having a positive impact on the direction of our global civilization than that of creating plugins for musicians.

Having said all this, I would love to be an alpha/beta tester of svng3; if I have any knowledge that could be of help, I would be honored to offer it. Needless to say, if I can pick a few cherries from the bountiful orchard of your own brain, I shall surely do so, given the opportunity.

By the way, GnuSpeech has a library of rules to modulate their own articulatory synthesizer. Maybe you could adapt them. Also, you are probably familiar with Perry Cook's stuff. Have you looked at his relevant patent?

PLEASE contact me via KVM's pm system (I just registered here today in order to facilitate contact with you. Would like to correspond via email.)

Bob H.

[xoxos]

xoxos.net > about
due to years of organised harassment i don't track down media people direct me to. viewing the cached copy of your front page doesn't say much about your method, so i don't know what you're looking for here. you may already know animals can speak english just fine, it's that "edified" humans do not hear. any international traveller may discern that birds use the same pitch inflections as the local humans. but civilised people want to think of themselves as the smart ones because they have stupid boxes with fast pictures for keeping dupes busy, ignorant and susceptible to authority by reference. i can hear them fine, must be the weed or decalcification or secret alien dna implants or something. i suggest weed. maybe i ought to come out with my own brand, makes you hear animals and program simple things.

anyway, syng3, experienced concatenators, i am confident you will be dropping by some time.

[spacedad]

what i meant was:i'd like to give it a try,even though time is scare atm.can't guarantee i'll be available at all times.but i can string a sentance together.and i'll buy it even if i was a beta tester.

[xoxos]

bob h. -
saw one of the messages on the se mail group. still don't know what you're doing, but for the things you mentioned, perhaps the SE SDK is an option - there are notes about using it with a couple of compilers around.. i have notes on setup with borland fclt, which is ancient, but simple to install and use. http://www.xoxos.net/sem/sem.html

gary -

[spacedad]

what? i can't even type anymore. (just realised my browser or computer has been slow and not catching all my type,bit odd behaviour,but it seems fine today.soz)

[xoxos]

there's a few weeks at least before any distribution. it looks like you were typing under duress from some kinda technological innovation there pardoner.

[kamalmanzukie]

ooh me me. i've coaxed some seriously suprising legibility out of syng 2, and after running it through impulse responses my own vocal resonance (don't ask how i got that), people have appeared to be appraising the back of my skull as if looking for any signs of coaxial cable.

[xoxos]

hopefully using a spark impulse not a starter pistol. i'm about 1/2 way to implementing some impulse based synthesis forms so the topic (impulses, not your vocal tract) does interest me (linear predictive coding otoh is "boring"), but i'm thinking this would have been done before sampling.

[seedpress]

Xoxos,

(Sorry for the delayed reply -- lot of colds going around here!)

On your web page I have read your disturbing accounts of gang stalking. I wish there were something I could do or say to ameliorate your resultant suffering. Though not on the scale of your experiences, my own life has had some striking parallels in this regard: I live in a very racist, culturally backward region of the Ozarks. Three years ago my small family and I were driven from our home by a gun-wielding white supremacist. We had black friends. We've had to relocate. It's a stupid, ugly story, and still not completely resolved; I won't dwell on it further here.

I only wish that people would not be so dismissive of those who have had experiences such as yours. At any rate, I'm amazed that in the midst of such turmoil you have been able to push out so much work!

Another of your comments: "i don't track down media people direct me to." Maybe someone else would be interested in the links. Beyond that, guess it's a moot point.

On to the technical stuff:

You commented: "saw one of the messages on the se mail group. still don't know what you're doing." (I'm not sure I do either! )

My questions on the SE user group are mostly related to one of two software projects. I'm basically just trying to figure out how SynthEdit works. It has a lot of poor documentation -- some is good, too. My questions were written late at night when I was tired. So, they weren't well thought out. I'll look at the SDK more.

The second thing I'm working on is a virtual vocal tract. The Manzara tract model (mentioned in my first post) has numerous feedback loops that are only one or two samples long (at 44,100). I have coded a plugin in Faust, to which a VST wrapper was added. One-sample feedback delays are possible in Faust. Without such short feedback loops my model breaks. I may even have to use the over-sampling fork of Faust to get what I need.

Unfortunately, Faust has limited GUI output. So I thought I might use SynthEdit to re-code my VST. When I started working with SynthEdit I was unaware of its sample-block-size, feedback loop limitation. So, given that limitation, SynthEdit won't work for the vocal tract model I 'm using.

Will have look at your own kltract, but as it's not open source, I'm not sure what can be learned there.

To segue to the philosophical:
You commented, ". . . you may already know animals can speak english just fine." Well, maybe in your world they do, but NOT in mine, nor in most other folks'. Animals do communicate, however. Some very well, in fact. Each animal I've lived with has helped me create our own special language. But people who visit me still can't understand my dogs or my cat.

If a dog had a vocal "instrument" that had the versatility of a parrot's, I am confident that they could then communicate with my friends in words they could both understand. Maybe you've heard of Alex the Parrot?
http://alexfoundation.org/
http://www.pbs.org/wgbh/nova/nature/pro ... -alex.html
If Alex can do it, then why not a dog? Here's some interesting info on a dog who was trained to recognize over 1000 English words:
http://www.pbs.org/wgbh/nova/nature/how-smart-dogs.html

Here's a quote from my web page "When animals themselves are finally able to speak out in a way that is both human-like and thought-provoking, then their new voices will also touch the consciousness of those humans who otherwise would never have been able to listen to what the animals have to say."

I'm doing this much as for all the humans who lack visionary insight, as for oppressed animals.

You wrote, “your front page doesn't say much about your method.”
My web page is meant to be merely a statement of goals, not an explanation of how the goals might be accomplished.

I am basically a hardware guy. But some of my project needs software, with which I constantly struggle – hence my brain-picking presence on forums.

I remember when I came out with the first consumer Alpha Wave Biofeedback device, everyone thought it was a crazy idea. Yet now there are so many companies making them that I've lost count. My experience is that even the most clever technical people rarely have true vision. That's why it took two guys in a bicycle shop to make the first flying machine, instead of the wealthy, well-equipped industrialists of the day.

Though we are still far down the mountain, we are also climbing steadily, and have the peak well in sight!

Good luck with your own work.

Bob H.
seedpress

[xoxos]

hmm... you know, SE can embed a vst, eg. old whatshisname's vocoder used to find its way into some popularly used plugs, a long time ago.. if you're only looking for gui.

haven't looked at the manzara model, if it's simple and definate, i can code a one-off module for SE and save you the trouble. of course, being a tormented and player resistant person means it's only going to happen if it's easy/explicit/won't require adjustment. you can judge if it's worth communicating.

[kamalmanzukie]

well, i've been using zynaptics unfilter demo that runs for fifteen minutes to extract them. it'll grab resonance off anything and turn it into an ir, a Stradivarius recording... my clean electric guitar signal, which can then be applied to a program like pling with great success. also hiding my head under blankets with a microphone making strange throat sounds trying to trick the plugin into thinking my voice is "the room' adding resonance. its great, but i'm pretty sure the some result or probably better could be obtained by picking out the desired frequencies in a program like spear. anyway i don't know that it would help for sampling, but just running the signal thru the ir seems to help lpc out quite a bit, especially if you are trying to get it to sound like your voice tone. if your sibilance is in there too, it seems to help with intelligibility. i also have some crazy notions about vowel formants and how they might possibly be used but really my understanding and methods are pretty hackish and amateur at this point. anyway, i'll see if i cant post a clip when i get home from work tomorrow.

[seedpress]

xoxos,

Late reply. Been sick for a few weeks (rare for me).

You wrote: "haven't looked at the manzara model." Here's a link:
http://www.gnu.org/software/gnuspeech/trm-write-up.pdf
He has a series of bi-directional delay lines connected with two-way scattering junctions and a three-way going to the nasal tract. You can use a one-multiply, three-add instead of the four-multiply, two-addition junction -- not that it matters much with CPU's these days. There are also lots of feedback loops.

I don't see how this can be done in synthedit because of the large sample-block size. The delay lines are only one or two samples long -- at 44,100. That's why I used Faust. It has a one sample delay for feedback loops. Even that may be too much.

I guess I need to bite the bullet-shit and learn C++. (Life is too short!)

[xoxos]

too short not to use c++!

excepting commands that integrate with the environment, the c++ for doing waveguides is only a few terms.. stuff you already know.