Robotic Voices

[gsilbers]

How do you make your Robotic Voices?

I use the text to speech in OSX and vocoder's like the new one from live as well as
a banshi talkbox2.

what outboard gear/software do you use or which ones are good for this?

[Absent1]

Well there is the Flame Talking Synth, but to be honest the novelty of it has worn off after a year for me.

It is pretty damn hard to make actual words out of it, however it does have alot pre-programed words as well as alot of sound effects.

[gsilbers]

oh , i remmebr that one from a computer music mag. looks nice though.

[gsilbers]

any other?

there was an nold korg voice sfx/ vocoder but dont rememebr which one. any ideas?

[idiotboy]

There's a long list of vocoders, both hardware and software. I've used the Electrix Warp Factory, the Korg Radias and the Roland V-Synth XT, and though it's not a vocoder, the Flame MIDI Talking Synth.

Most people say the vintage vocoders made by Roland, Moog and others are among the best.

[gsilbers]

cool. anything else thats not a vocoder?

[ElMosca]

Originally Posted by gsilbers

How do you make your Robotic Voices?

I use the text to speech in OSX [...]

To answer your question: I mostly use the microKORG vocoder and the Reason Vocoder.

Now to ask my question: how do you capture speech from the OSX speech engine?

I'm using Audio Hijack, and I can't find the speech synth stream to hijack/record. I tried hijacking the Finder as well as TextEdit (where the text I want the computer to "speak" is typed).

Suggestions???

Thanks,
++aldo

[Yoozer]

Originally Posted by gsilbers

any other?

there was an nold korg voice sfx/ vocoder but dont rememebr which one. any ideas?

Korg VC 10.

Anyway, a Nord Micromodular does a good job too. Today, I've played around with Ableton's vocoder which is pretty decent. No Eiosis ELS or Virsyn Matrix but quite capable and far better than Prosoniq Orange.

[space2012]

EIOSIS Vocoder has some vintage robotic sounds.
starplugs vocoder
waldorf vocoder,
novation supernova vocoder
etc...
i dont like the tc fireworks vocoder.

vocoders can be done with modular synths.
just a bunch of band pass filters like a graphic eq, each with a envelope follower, controlling bandpass filters with vca of another signal.

SignalA-->Bandpass1A-->Env.Follower1-->VCA-->Bandpass1B-->SignalB
A-->Bandpass2A-->Env.Follower2-->VCA-->Bandpass2B-->B
A-->Bandpass3A-->Env.Follower3-->VCA-->Bandpass3B-->B
A-->Bandpass4A-->Env.Follower4-->VCA-->Bandpass4B-->B
A-->Bandpass5A-->Env.Follower5-->VCA-->Bandpass5B-->B
etc...
even you can do weird stuff like cross mod, change the Q in the filters, do inverse connections Env1-->Bandpass5, change the frequency of the filters, etc...

[space2012]

[nikseven]

i use a Roland SVC-350, and i highly recommend it. beautiful sounding vocoder. in the past i have also used an Electrix WarpFactory quite a bit, and it is very clear and intelligible, porbably the best digital vocoder i've tried, but it does definitely lack the warmth and character of the analog vocoders. for awhile i was running both the SVC and the Electrix simultaneously, but i've sinced stopped using the Electrix altogether, and only use the SVC. i also have a Koeg VC-10 which is pretty cool, but i find it to be pretty volatile and inconsistent, and it takes a lot more work for me to get a sound i'm happy with compared to the instant gratification of the SVC-350.

[Jim Stout]

What about the Korg DVP-1?

As far a I know it's the only 12bit vocal tweak box ever made.

[misnomer]

You could use 3 Bandpass Filters, look on page 9, there you see the specific frequencys for each vowel:

http://www.doepfer.de/a100_man/A127_man.pdf

I do this with a Technosaurus Triple Resonator..

[gsilbers]

Originally Posted by ElMosca

To answer your question: I mostly use the microKORG vocoder and the Reason Vocoder.

Now to ask my question: how do you capture speech from the OSX speech engine?

I'm using Audio Hijack, and I can't find the speech synth stream to hijack/record. I tried hijacking the Finder as well as TextEdit (where the text I want the computer to "speak" is typed).

Suggestions???

Thanks,
++aldo

there are also free softwares u can use. there used to be one called voicebox which you could change pitch also. very cool but its discontinued. (osx)

btw, i tried looking into real voicebox, like the ones from emphysema patients but didn't find much,

[killedaway]

the Electrix WarpFactory is my de-facto, go-to robot machine.

[cane creek]

I sold my Electrix WarpFactory on ebay a few years back i wonder if it was one of you guys who bought it , i bought it new for £140 used it for a few years then sold it on ebay for £140 so it kept its value

Ive had good results in the past with Native instruments Vokator however NI dont support it no more.

At the moment i use my MicroKorg XL which i get great results , if its something i want to add quick to a sample ill use logics own vocoder.

[msl]

I use an Alesis MetaVox, not really a vocoder, but lots of fun none the less. They can be had cheep second hand, got mine for €40. I'll second the new microKorg XL it has a really nice vocoder on it. I've also got some good results using the one in Reason, but its PITA cause you have to prerecord your audio.

As for the question about system audio on the mac, its easy just set Hijack to record from Application, then select Text Edit. I use the same setup, works fine. You might need to route your audio out to the built in speakers, if you have it set to your audio interface.


.

[cane creek]

[rockmanrock]

Originally Posted by killedaway

the Electrix WarpFactory is my de-facto, go-to robot machine.

I can see the appeal:

YouTube - Electrix Vocoder


The cheap Zooms aren't bad either:

YouTube - Zoom Studio 1204 Demo

[jtothap]

I use a Nord Modular which has an awesome sounding vocoder, but I love it for really vocally sounding vocoding.

For robots I havent found anything better than the EIOSIS Vocoder. I've heard so many vocoders in tracks that you can barely understand. The EIOSIS vocoder is so damn clear, and really dirty and noisy if you so desire.

[aeonlux]

As has been mentioned, Apple's Text-to-Speech routines can generate some nice robotic voices. This is especially true when considering some of the older voice models from MacOS7 that did not migrate to MacOSX - they are more primitive and robotic-sounding.

The Elektron MonoMachine SFX-6 has a vocal model that can be made to generate some robotic (and otherwise) speech. For longer phrases I prefer to use other methods as I am lazy and this method requires some work.

The Korg OASYS PCI has a number of vocal physical models that can be modulated so as to make them talk. That said, as it concerns full speech I prefer other methods because driving an algorithmic physical model to do full speech is not easy. The OASYS PCI also has an excellent vocoder courtesy of Dan Philips.

The Yamaha FS1R can generate excellent-quality robotic speech when using its formant sequencer. The voiced and unvoiced operator combination allows you to carefully balance the degree of synthiness and breath-noise in your robotic model.

The E-mu Ultra series of samplers with RFX card allows you to use a glottal noise or synth sample as the source sound, or use live input from a mic, then process with a vocal formant sweep filter, then pass this audio into the RFX card where it can be processed by up to 4 simultaneous vocoders, even in stereo. Audio-to-trigger means you only need to use the mic, with no key playing, to trigger the processing chain.

The Clavia Nord Modular can be made to synthesize some basic speech, but its real treat is its vocoder - classic sound and the means to cross-up the bandpass filter relationships means that all manner of odd and robotic voices are just a few mouse-clicks away. They are especially effective when using other modules to shape the output.

Digidesign ProTools running WaveMechanics/SoundToys PurePitch, PitchBlender, and TimeBlender is an excellent suite by which to process audio with inflection control, formant shift, and reharmonization so as to create robotic and otherwise voices.

Last, but certainly not least, the Eventide H8000FW Ultra-Harmonizer is a demon as it concerns creating characterful voicing. You are provided with both the tools and the DSP power to create processing engines for all manner of robotic (and otherwise) voices, including ensemble shifting so as to create groups and crowds via pitch, formant, and time delay, all with environmental reverb and ambience.


cheers,
Ian

[Delta Heavy]

which of the cheaper options would be closest to the roland svc350?

[killedaway]

okay, this is actually a pretty awful example of a vocoder, at least in terms of intelligibility, but the comments are not to be missed!

YouTube - 空(vocoder test:ZOOM STUDIO 1201)

[aeonlux]

Originally Posted by Delta Heavy

which of the cheaper options would be closest to the roland svc350?

Clavia Nord MicroModular


cheers,
Ian

[aeonlux]

For great results in the least amount of time, I think and feel it is best to combine real speech or synthetic speech with vocal-oriented post-processing, and perhaps a bit of other oddness, e.g., ring-mod, mixed in at low level in parallel.

So:

Apple Text-to-Speech --or-- real human voice --> (Vocoder) --or-- (Formant/PitchCorrect) --> (Synthy Mangling) --> (Added parallel Synth tones)


cheers,
Ian

[crufty]

vsynth has nice vocoding, maybe not so much robotic voices though.

[pr0gr4m]

Vocoders give you a robot sound but not the robot pronounciation. I used to have this old text-to-speech program (I'll have to see if I can dig that up) that sounded so not human. I would sample words/phrases from it and then play them back through a vocoder. That would give me the best robot voice.

[aeonlux]

Using a buffer overrun/s-t-u-t-t-e-r processor, especially when combined with the right compression and bit reduction, can create the sound of a robot that is digitally-fried or breaking down.

Using a pitch correction tool to remove all inflection when not using a vocoder will help to remove emotional expression from the voice, and make it sound less human.

Another thing to try is when you have your phrase completed, ReCycle it into its parts and then use some method to alter/mangle the timing and cadence. This also helps to sound less human and much more mechanical. Subtlety is key when using this technique, especially when the voice has already been vocoded or pitch-corrected.


cheers,
Ian

[Yoozer]

Furthermore, use a carrier with non-standard tuning or say, 50% oscillator pitch tracking. It's far easier to mimic regular speech, because speech doesn't use semitones only.

[Jim Stout]

The V-synth GT or XT would be the closest to the the VP They actually sampled the VP waves in the Vocal designer.

just fyi