Korg Forums

I have played around with every parameter on this R3's vocoder and cannot get anything remotely clear and intelligible to come out of it. Any words of wisdom? or should I just sell this thing?

_ I really want to make a deep, powerful "voice of god" type of thing. Or even some voices like Midnight Star used. - electro funk band, sang "no parking on the dance floor"

bradrm250 wrote:I have played around with every parameter on this R3's vocoder and cannot get anything remotely clear and intelligible to come out of it. Any words of wisdom? or should I just sell this thing?

_ I really want to make a deep, powerful "voice of god" type of thing. Or even some voices like Midnight Star used. - electro funk band, sang "no parking on the dance floor"

Hi, well I was asking the same questions and decided to go deeper researching the circuit board inside. Because the R3 service manuals are not available it was a bit more complicated to perform a solution. Audio quality only affects DAC output, not the vocoder ADC input. Some users say that R3 vocoder is better as legacy Korg VC-10. Keep in mind that maybe you need to mix an external audio source for example a keyboard instead of using R3 factory waveforms as carrier for the vocoder effect... maybe there are more appropriate for DJ and "robot" style effects.

On my blog I share an article about the findings on R3 circuit design and the solution that I do.

http://www.alfredoblogspage.blogspot.co ... ality.html

Here's the final R3 audio quality after circuit bend

<iframe width="100%" height="166" scrolling="no" frameborder="no" src="http://w.soundcloud.com/player/?url=htt ... "></iframe>

Here a vocoder test. First an introduction about me in direct audio patch 8 and then I switched from patch 1 to 7 different vocoders effects. Vocoders on Korg R3 are excellent.
<iframe width="100%" height="166" scrolling="no" frameborder="no" src="http://w.soundcloud.com/player/?url=htt ... "></iframe>

And to end a fantastic legacy advertisement about Korg VC-10
<iframe width="420" height="315" src="http://www.youtube.com/embed/37jKhl7BBvM" frameborder="0"></iframe>

Again the DSP processor inside R3 is very powerful, is the same as any DSP used for a vocalizer or effects unit. This means that changing the DSP patches the R3 would be capable to perform harmonizing in the same way as a studio vocalist, unfortunately this modification can´t be done....

Regards

bradrm250 wrote:I have played around with every parameter on this R3's vocoder and cannot get anything remotely clear and intelligible to come out of it. Any words of wisdom? or should I just sell this thing?

_ I really want to make a deep, powerful "voice of god" type of thing. Or even some voices like Midnight Star used. - electro funk band, sang "no parking on the dance floor"

Hi Brad,
even if I don't know what you mean with "voice of god" (i have also heard from devils or hells voices) but he song you quoted (no parking..) uses a vocoder playing lower notes on the background. Such effect is also used on Phill Colins "in the air tonight" or Saga's "carefull where you step". Another classic vocoder effect can be heard on YMO's "behind the mask" or even more actual the background on Madonna's "masterpiece".

To get good results from a vocoder there are some basics you should know. If you take a look at my webpage (in my signature) you will find some hints as well as vocoder patches for the R3. Maybe you try the one called "Saga" which is possibly the sort of effect you are looking for.

And finally, the R3dias vocoders are one of the best ones you will find in VA synths actually even if the presets are not the best ones!

This is just really frustrating... I listen to all these old Electro Funk bands and tv shows like Transformers and you can hear the words clear as glass.. and everything I try to make with these vocoders sounds like mud with no clear syllables.

Your patches sound much better than mine... good job! But I still need a bit more clarity on the words.

My song will have little to no un-processed vocals, so they need to be clear like the ones from the 80s.

Am I doomed to spend $1000+ on a Roland SVC 350?

Have you ever tried making your own vocoder patch?

Do you understand how the vocoder itself functions- and how to take advantage of this?

Use pulse waves for clarity in words, saws to thicken the sound. Try to keep everything somewhat in phase, and feel free to dial in some noise for extra carrier functionality. Perhaps add some drive; run it through an open resonant lowpass and perhaps a high-pass right after to clear out some muddiness. Waveshape a little drive, compress the entire patch, use minimal modulation effects for clarity. Make sure your envelopes aren't doing anything funky unless you intend to hit a note with each syllable you speak/sing. It doesn't matter if the carrier timbre sounds like crap au naturale- it matters how it sounds when being used with the vocoder. Make sure to adjust the vocoder settings as well; shift your formant to taste, filter what you need to, and continually adjust for clarity.

Finally- of course a vocoder patch with the actual voice bled in or layered over will sound more clear!

Also, realize that not every robotic vocal effect is created via a vocoder- in the modern day, programs like Antares' Auto-tune and Celemony's Melodyne are frequently used to create perfect pitch, tweak formants and remove vocal vibrato. Linear-predictive coding, a precursor to some types of vocoding and a technology used in cellular phones, can also be interesting, with programs such as Sonic Charge's Bitspeek making for creative vocal timbres. Merely downsampling a voice, playing it back at a slower/faster speed or pitch shifting it while constraining rhythmic values can also result in a robotic sound.

http://soniccharge.com/bitspeek
http://www.celemony.com/cms/ (I heavily recommend acquiring a copy of this for easily manipulating rhythm and pitch of any audio)
http://www.antarestech.com/products/auto-tune-evo.shtml

Example Patch
http://dl.dropbox.com/u/1715217/shesellsseashells.m4a (the recording clips a little, my bad.)

Unfortunately I can't get the patch to you due to my current setup, but that was made by using Mono 2, three unison with 2 detune and a bit of spread (a bit too much, I'd use less spread). Oscillator one was a saw wave in unison mode with no detune or phase offset; oscillator two was a saw in sync mode 9 semitones up. There was a lot of noise added (almost full) and punch turned to around 70.

Filter 1 was a bandpass filter set to near-thru; with resonance turned way up, keytracking set to +1 and the cutoff tuned so that the resonance was the same pitch as the note. It was then run in serial to filter 2, a comb filter; Cutoff was at 60 with +1 keytracking, and resonance all the way up. Pre-filter drive was applied, about 1/3 of the way.

The volume envelope had the slightest bit of release, full sustain. The EQ was used to gain a lot around 800-1000 hertz and 1800-2400 hertz.

The vocoder was enabled; FcOffset was set to a fifth (+7), vocoder resonance was around 20-30, and E.F. Sensitivity was very little.[/b]

bradrm250 wrote: ... Am I doomed to spend $1000+ on a Roland SVC 350?

No.
Even if you would get an EMS or whatever else top of the ranking vocoders, you wouldn't be satisfied until you get the essential basics of vocoding. Besides the carrier that as always stated, should be simple and harsh (our vocal chord is a single pulse oscillator), the main issue is the formant (or vocalization). To achieve the best vocalization from speech, you need to optimize it using EQ and compression. If you take a recording of a news speaker or radio narrator for vocalization you will see the difference at the output compared to your own voice.

Other important factor is what do you expect to achieve with the vocoder: use it as background to your own voice or use it as the lead voice.

In case you want it as background, then your natural voice is what makes it intelligible.

If you want it to be the lead voice, then you have to improve your vocalization, try to maintain a constant pitch (speaking like a robot instead of singing) to avoid fluctuations in the formant spectrum and hence in the result. In this case it is also very important to realize that the background instrumentation must be complementary to the vocoder spectrum to avoid drowning or overlapping the vocoded part.

Until you get a real good vocalization (which also includes your mike!) your attempts will always sound like mud, no matter what vocoder you use. This is as natural as with your own voice, which will sound like mud too if you are eating smashed potatoes. Our natural speech quality may also be poor due the natural caracteristic of our vocal apparatus which is why not everey one of us may be a good singer or narrator (like me).

On page 59 of manual is explained. The parameter to change is DirectLv (direct level) that works similar as the dry-wet feature on a effects unit. This will adjust the quantity of vocoder effects versus direct voice from mic, and in this way you can adjust "clarity" on the speak with not so much vocoding effect.
Also try to check other vocoder parameters, R3 vocoder is very flexible. Don´t forget that you need to source an external carrier if you want a custom or special vocoder sound, avoiding to use the inside waveforms.
R3 vocoder is the best option today comparing analog units, the power of a DSP to do this job is more efficient and precise as any legacy analog gear. This is the reason why I'm very interested to transform my R3 to a studio vocalizer...
If someone can provide some tips on how to modify the firmware eliminating some thing for example the demos songs and use this memory space for harmony-vocalizing programs effect would be great.

Shift+8 - Off - 8
Write - Write - Write