Standalone Vowel Generator

[Unkie Al]

Once again I need to be pointed in the right direction here...

I think we have all seen the "Delay Lama" VST plug-in, where a little cartooney monk is chanting vowel sounds from EEE to OOO, with stops at AH, AY, IH and others along the way.

Is there a source of info to pull this off in hardware? A VCO feeding a formant filter, with both pitch and timbre steered by a cheapie joystick?

If not a dedicated device, how about tips on sites addressing formant/vowel filtering so I can take it from there? Thanks again as always.


UA in DC

[yusynth]

On my site, yusynth.net, in the archive section you'll find the PDF of this paper
http://yusynth.net/archives/index.html
TOPIC
Voice formant synthesizer
TITLE
The Chatterbox : A simple speech synthesizer for demonstration and amusement
AUTHOR
Ian Witten and Peter Madam
JOURNAL
Wireless Word december 1978
DOWNLOAD
(PDF 1.2Mb)
COMMENTS

Here is a direct link to the file :
http://yusynth.net/archives/WirelessWorld/Chatterbox-1976.pdf

Enjoy it

[Unkie Al]

Huh! Pretty freekin' cool. Thanx!

UA in DC

[julianw]

You should check out

http://www.analoguehaven.com/flame/miditalkingsynth/

Sample1
Sample2
Sample3

A very similar DIY implementation using the same chip (Speakjet) using the midibox platform is here:

http://www.ucapps.de/index.html

Click the scroll down the frame on the right and click the link to MIDIbox SpeakJet under 'User Projects'

[ickystay]

About powering the Chatterbox-1976.pdf, the article says "Two PP3 batteries provide an internal power supply". Anyone know if it means go with 18vdc, or is it +/-9vdc?

[Luka]

how about speakjet chip

they have a section on it over at ucapps i belive

[yusynth]

[ickystay]

Thanks for clearing that up. That was my guess, but I did wonder about running cmos chips that high.

I have a 4 pot joystick salvaged from an old quadraphonic home stereo reserved for this.

Thanks also for this pdf and all the other cool stuff on your site!

[yusynth]

[fluxmonkey]

yves--

have you built this? any experiences to share?

b

[yusynth]

No I haven't built it...

[bugbrand]

This is a very interesting project and I am sure it could provide a very good basis for making something A.C.E. - very valuable source of info f'sure.

I remember trying it out a few years ago, but hadn't clicked that you needed a bipolar supply.. oops!.

I don't see the need for:
1) bipolar supply for the oscillator/noise CMOS chips really. But it IS currently necessary for the 4016 (though that could be worked around)
2) The noise source seems a bit o-t-t --- I don't really see the need for it to be gated - its being gated already by the 4016s later... Oh, well as it stands you could easily make it vari-speed.
3) Finding a dual-gang joystick is tricky! I've been thinking to implement dual vactrols for the formant filtering parts.
4) Formant Filtering could be a very nice stand-alone effect! -- perhaps very similar to Fonik's PS3100 resonator clone - Hmm, must look back at the schem for that..

I'm very tempted to get sidetracked by playing with this circuit!

[frijitz]

I don't believe anyone has mentioned Jim Patchell's vocal filter:
http://www.oldcrows.net/~patchell/synthmodules/vocalfilter.html

[mandrigora]

479 components...That's..a...lot...of...solder burn blisters....

[blue hell]

the component layout on the PCB looked quite spectacular indeed, some nice samples on the site BTW.

[andrewF]

This has been a goal of mine for a long time.
I have experimented with SPO256 chips in various set-ups, the best was using part of a Milton engine board, but I could never get a decent sound out of the output - I`m not that fussy but it sounds like crap - thin and weedy.

I have Jim Patchell`s Vocal Filter, it is very good, but a major build and i think he is out of PCBs, it would be an awesome veroboard project
but to be honest I have never got it to output the `monks chanting` type sound, but there are a lot of knobs to tweak so it must be in there somewhere.

I have never built the chatterbox, despite it being around the net for a long time, never heard of anyone who has.

Recently, I discovered an analogue implementation of the VOSIM in Electronotes and am planning to build that. I asked about it on the Nord discussion board and some people kindly posted some samples of the VOSIM setup on a Nord. The analogue version works by hard-synching the VCO and (as I understand it) fading out the sound as the VCO is `switched` off.
Guess you need EN for that, or maybe Bernie could post the article. It is not one of his designs though.

The Speakjet (and its newer version) are really programmed PIC chips, you need a serial input to control them. I have one of these too but it always seemed to much bother. guess I`m waiting for someone to put out a PCB One guy Lorin(?) produced a PCB and planned to make more for selling but he seems to have moved on to other projects.

Some of the best vowel sounds I have got come from the CGS Steiner filter! especially if you have another filter that can go sub-audio and control the CV signal driving the VCO. It can sound very realistic and quite disturbing at times.

[blue hell]

[andrewF]

I have had good results with Milton sequencer ->filter ->VCO ->Steiner
or with an EG instead of the Milton.

Sorry got no samples, I am in Japan and my toys are in Australia.

[Randaleem]

[doctorvague]

This has some pertinent info:
http://www.analoguehaven.com/doepfer/a127/manual.pdf

Check out page 9.

cheers
Phil

[doctorvague]

a couple of follow-up ideas to the above link...
It seems that 3 is the magic number (or a minimum of 3) formants to produce human vowel-like sounds. So 1 joysick would not cover all 3 parameters. Also the range of frequencies is specific and fairly narrow.

Some things I would think would be important are:

1 Bandwidth - I should think you'd want a variable bandwidth for each filter band, and you'd want a really narrow bandwidth at least available in that range. You probably wouldn't need a wide bandwidth setting, so you're looking at "really narrow" to "pretty narrow" (to be exact

2 scaling - if each band could be scaled to only it's useful frequencies, you'd hve a much more playable instrument, in that it would stay in vocal formant range. E.G.- The useful range of your lowest formant might be 250-400 (just made-up numbers, not a real example). So having upper and lower scaling and limiting your knob/slider/joystick to ONLY those freqs gives you a completely usable range on your knob- rather than only needing 9:00 - 9:30 or something (making it difficult to tweak and making it very easy to go OUT of human vocal formant range.)

The other thing to think about with the kind of scaling control I'm talking about is you could really tweak it to go ONLY from "A" to "I" for instance across your pot range(s) and pick your vowel sounds - IOW narrow down your ranges to the exact vowels you want to use. Scaling incoming CV's makes sense for the same reasons above. E.G. - an envelope scaled to only cover "A" to "I" as above.

3 control - if it were me, I'd want independent control over all 3 freqs, like 3 knobs or 3 sliders (and 3 CV's). Or maybe a knob and a joystick and be able to assign the bands to the controllers? This part is subjective obviously but the main point is I would want control over all 3 and not just 2 with a joystick.

All these ideas are just for the actual vocal fromant filter only, the speakjet for example being an entirely different approach. In fact it's much more akin to sampling if I understand it correctly.

Just some thoughts for the idea-heap
Happy ooh'ing and aah'ing
Phil

[bugbrand]

Yeah, the Chatterbox is strange in that it is just two filters and they're in serial rather than parallel. Actually, the PS Resonator board Fonik did IS three bandpasses (in parallel) and in use I've really liked the vocal-sounds available through it.

[Randaleem]

Hi Phil,

Thank you for the Analog Haven link! And your Good points.

EDIT: Bulk of message moved to new Chatterbox PCB thread.

Kind regards, Randal

[Rykhaard]

In my 'Vowel Filter' that I started building in the mid 90's, I had set up trimmer resistors for each of 3 formants for 5 different vowels.

Each of the vowels that I had breadboarded before it, with a Pulse wave input to them sounded HUGELY accurate, as far as a male voice goes.

I never DID complete the module, BUT - I still have it.

I'm now considering measuring the value of each trimmer to replace them with resistor pairs (or triplets), to free up the 30 trimmers.

The circuits used were high resonance 2 pole BPF's, out of Electronotes.

I just may go back TO this module to complete it, to finally satisfy my vocal filter desires.

In the original idea, the vowel for each of the formants were to be selected by rotary switch. I'm going to rethink it and see about VC switching as a 2nd possibility as well. (There was no Fc adjustment for any of them. Just hard tuned, to the desired frequencies.)

Yet ANOTHER thing, placed onto my backlog of modules to get completed. (Bumped back up to 9, with the Klee boards arriving. )

[doctorvague]