Argotlunar - a novel method of voice scrambling

[Dr Jeffers]

I was inspired to look further at voice scrambling techniques after trying Lance's method, and seeing how effective it was. I investigated Granular Synthesis, and found a (free) VST plugin whose functionality could scramble speech - called Argotlunar.

Argotlunar is a real-time delay-line granulator VST plugin, that I installed into Wavepad. It breaks up an incoming audio stream into short samples (Grains). Each grain can have random settings of amplitude, panning, duration, delay, pitch, glissando, filter and envelope applied to it. The output of all grains is mixed, and sent to the main output. The main output can be fed back into the main input.

For voice scrambling, only basic settings are needed:
"Grains": to 1.
"Delay" (maximum random period span range): 0 to 5,000mS.
"Dur" (duration of each grain): set to fixed value (no span) of 235mS - or whatever suits.
Envelope shape set to "RCB" (raised cosine bell).
"Sustain" (rise/fall shaping of RCB curve) 50% rotation (sounds good).
Other settings as per screenshot.
Presets above can be saved and recalled.

It produces a pleasant sounding scramble, and I have enclosed an output sample file, as well as comparison files from EVPMaker (235mS with Z-cross ticked) and Lance's method (1,244 files of 235mS, played randomly in Audacious). I did not remove the silences in between words in the source file.

A major limitation of Argotlunar for scrambling is the 5 sec delay span within which grains are selected, shaped and played. This delay window follows the play-head in real time, so as new grains enter the span, old grains are discarded. This means at any one time (for 235mS) there is a rolling total of 21 grains available for selection. In listening to the output file, there some repetitions heard.

Ive only played with this concept with a free plugin, and I know more advanced (paid-for) products are available with better randomization.

I put it out here so perhaps others might want to take it further.

Download is here...
https://mourednik.github.io/argotlunar/

JEFF

Argotlunar 235mS.mp3 EvpMaker 235mS.mp3 Lances Method 235mS.mp3

Edited June 2, 2021 by Dr Jeffers

[Lance Reed]

Thanks for sharing Jeff, I will have a good read of this post and experiment later in the day,
speak soon, Lance 

[Shaye]

Thanks for sharing Jeff!    I’m not familiar with adding plugins to Wavepad, and my first attempt at installing  Argotlunar was unsuccessful.  I downloaded the file, added it to a designated folder and extracted the file.  Went to VST in Wavepad to install the plug-in, but was told no plugins were found in the folder.  Not sure what I’m doing wrong, but will continue to experiment.    

[Lance Reed]

5 hours ago, Shaye said:

Thanks for sharing Jeff!    I’m not familiar with adding plugins to Wavepad, and my first attempt at installing  Argotlunar was unsuccessful.  I downloaded the file, added it to a designated folder and extracted the file.  Went to VST in Wavepad to install the plug-in, but was told no plugins were found in the folder.  Not sure what I’m doing wrong, but will continue to experiment.    

Hi Shaye,  try downloading the win 32  instead of the win 64, I had the same problem when I downloaded the 64, but the 32 worked for me,
Take care, Lance

[Dr Jeffers]

Yes, me too, the x64 version did not work for me either.

Just to clarify: the path to the VST plugins is set in tools/options/VSTs. Add a path to your Argotlunar files there. Restart. Load a sound file, goto Effects/VST Plugins, select Argotlunar from drop down menu. I get an initial corruption appearing that shows a "NO VST Selected" message in the middle of the Argotlunar console, but this is removed by maximising the console window, then restoring the window to normal size.

Good luck!

JEFF

Edited June 3, 2021 by Dr Jeffers

[Dr Jeffers]

Just had an idea that Ive just tried, and that was to Argotlunate the EVPMaker file I had already recorded. I think this is good, and did solve the occasional repeat problem of course, and the shaping of grains helps the EVPMaker file to be more pleasant and natural, but I think the secret here is to EVPMake the original source file at a greater size - say 350mS or larger, so Argotlunating the file at 235mS creates more consistently sounding grains.

An another thing I tried with Lance's method was to try and reduce the clicking due to the conjoined playing of sharply split files. Got some benefit by enabling cross fade in Audacious. See pic.

JEFF

Edited June 3, 2021 by Dr Jeffers

[Lance Reed]

It's a great bit of software Jeff, I will try that tomorrow what you suggested.
Take care, Lance

[Shaye]

Much thanks Lance and Jeff!  Downloading the win 32 instead of the win 64 solved my problem

[Lance Reed]

Hi guys, 
I Run a session using  Argotlunating this morning, had a few EVP through, It granulates the voice really well.........    But I had a thought after the session, to run the raw voice through the Argotlunating software to granulate it, then segment that saved audio to 225Ms and run a live session through Audacious using the mixed segments.  Will let you if I get any results in the morning,
take care, Lance
PS, here's one from this morning session just using Argotlunar, I hear E.V.P being spoke (I am using David Fontana voice for this session)

EVP.wav

[Andres Ramos]

Great proceedings. If I get you right Argolunatic also provides feedback. In that case it would be a more or less complete replacement for EVP Maker.

[Andres Ramos]

I could extract some nice samplesfrom the Argolunatic sequence.

You're a cute wick.mp3

To innovates.mp3

Think we heard it Edison.mp3

Or it's FEG.mp3

It must be Robert.mp3

Britains are all waverish.mp3

Come back and cooperate.mp3

It must be Robert.mp3

Do you ever entered this.mp3

[Dr Jeffers]

Yes, good samples you guys have obtained.

Well Andres, Argotlunate does have a major limitation in that it only randomizes within a rolling 5 second window that follows the play-head, so eventually the whole file does produce a scrambled audio sequence, but in a progressive manner as the file plays through in real time. So for a 235mS sample taken from a 5 rolling second window, its a limited random choice of 1 in 21 grains. EVPMaker produces its scramble from using the whole file span, so choice of 1 in 2,000 or so, depending on file length - so much greater randomisation.  But in saying this .....

I like Lance's idea of using an Argotlunate scrambled sequence, splitting it into segments, and then using Audacious to play the segments.

Im glad you guys have picked this up and are experimenting.

JEFF

[Andres Ramos]

2 hours ago, Dr Jeffers said:

Yes, good samples you guys have obtained.

Well Andres, Argotlunate does have a major limitation in that it only randomizes within a rolling 5 second window that follows the play-head, so eventually the whole file does produce a scrambled audio sequence, but in a progressive manner as the file plays through in real time. So for a 235mS sample taken from a 5 rolling second window, its a limited random choice of 1 in 21 grains. EVPMaker produces its scramble from using the whole file span, so choice of 1 in 2,000 or so, depending on file length - so much greater randomisation.  But in saying this .....

I like Lance's idea of using an Argotlunate scrambled sequence, splitting it into segments, and then using Audacious to play the segments.

Im glad you guys have picked this up and are experimenting.

JEFF

Ah yes, now I understand it. This could yield in randomizing speech but the outcome could reflect the same context as the source.

[Lance Reed]

Very nice responses Andres, sounds great!
Thanks for the info Jeff, makes sense.
Take care, Lance
 

[Dr Jeffers]

Lance, I tried your idea of splitting up the Argotlunatic file, and then playing the segments randomly in Audacious.

This worked very well. I used "Split At Silences" in WavePad, and set the parameters to suit the source file noise floor, etc.

This method preserved the pleasant shaped envelopes of the Argotlunatic grains, and gave a good randomization (1 in 364 for a 2min source file length).

Playback in Audacious did not need the crossfade effect applied, as there was no clicking. I think this is as good as it gets for using this plugin.

Edited June 7, 2021 by Dr Jeffers

[Lance Reed]

Thanks Jeff, I will try this what you suggested, for tomorrow morning contact session,

thanks again, Lance

[Lance Reed]

Hi Jeff, could you take a screen shot of the process, wave pad keep telling me that it can't find no spaces to split the file?
Thanks Lance

[Dr Jeffers]

Hi Lance,

Here is tis....

[ I did not delete the spaces between the grains of the recorded Argotlunatic file, as there would be no spaces to detect using this process ]

JEFF

Edited June 8, 2021 by Dr Jeffers

[Lance Reed]

Thanks Jeff.

[Dr Jeffers]

I did recording runs using the different methods weve discussed.

I got responses from all methods, but the best was splitting the original source file and playing that in Audacious. This may be due to the more abrupt nature of produced segments using this method, compared to the softer nature of the grains. Probably not conclusive. Will be interesting to hear results from others.

Stop_shouting_of _emaralds_there.mp3 This_gate_add_to_us_if_he_and_others_approve.mp3 We_need_magnets_and hoses_here.mp3

Edited June 9, 2021 by Dr Jeffers

[Lance Reed]

Beautiful contact Jeff! Thanks for sharing.
I have also found so far that the segmenting of the source file at 225Ms then running it through Audacious produces the best results for me.  I feel that the scrambled voice doesn't need to be fast, as long as it's fragmented and gibberish in nature, I feel the other side will do the rest.  
Here's some from a recent session, take care, Lance

 

Telsa.wav

[Lance Reed]

Girlfriend_ Sister.wav

[Lance Reed]

we began set voice over tape.wav

Edited June 9, 2021 by Lance Reed

[Lance Reed]

voice complication.wav

[Lance Reed]

Thanks Jeff,
I.T.C is sure a very intriguing subject, I rack my brains day and night thinking, how can I improve this and that etc. lol 

Have you seen the new voices through light that Sonia and her team are working on? Looks good, but I am no engineer to build things, but what I can see is happening , is that voice is transferred through light  waves and then converted back to voice?? Looks good.
Take care, Lance  

[Andres Ramos]

On 6/7/2021 at 3:05 AM, Dr Jeffers said:

Lance, I tried your idea of splitting up the Argotlunatic file, and then playing the segments randomly in Audacious.

This worked very well. I used "Split At Silences" in WavePad, and set the parameters to suit the source file noise floor, etc.

This method preserved the pleasant shaped envelopes of the Argotlunatic grains, and gave a good randomization (1 in 364 for a 2min source file length).

Playback in Audacious did not need the crossfade effect applied, as there was no clicking. I think this is as good as it gets for using this plugin.

What a cool idea!

[Lance Reed]

Thanks for the info Jeff, I would love to build the first faze of this model Sonia tried, but I am no good with electrical stuff  

[Keith J. Clark]

On 6/3/2021 at 9:43 AM, Shaye said:

Thanks for sharing Jeff!    I’m not familiar with adding plugins to Wavepad, and my first attempt at installing  Argotlunar was unsuccessful.  I downloaded the file, added it to a designated folder and extracted the file.  Went to VST in Wavepad to install the plug-in, but was told no plugins were found in the folder.  Not sure what I’m doing wrong, but will continue to experiment.    

Hi Shaye, I demonstrate how to add vst plugins in vsthost here: (maybe it will help?

I hope we'll get to teach this more.

Jeff, this is a very well-thought post! I like how you demonstrates comparisons.