Synthesis Magic - Creating a convincing choir derived from noise

A couple years ago I became fascinated with the technique of recursively processing an audio signal through a resonant polyphonic filter to hone in on a single frequency. I first saw this demonstrated by Jakub Ciupinski at his Supercharge your VCF episode of VCV Rack Hacks.

I invested a lot of time investigating what filters worked best, and I discovered that the Squinky Labs F2 filter does an amazing job of converting noise into a convincing emulation of a choral voice. (especially when V/Oct pitch changes are slewed a bit, and a simple reverb is applied). I tested many filters. (Someday I need to publish my testbed patch). Many filters are not able to hone in on individual frequencies well enough. Others are too perfect, and the result is flat and boring. But the F2 has just enough extra harmonics to make it sound like a human voice.

Wiring up the recursive filter used to be a bit of a pain. But since the release of my Venom plugin, the Recurse and Recurse Stereo modules make polyphonic recursive processing trivial to implement. In my patch below I also use some Venom mixer expanders to fade the sound in and out. The mixer expanders have not been released yet. But they should be released to the library within the next month.

The other necessary component to my end result is the hopefully soon to be released Squinktronix Harmony module that is able to create beautiful harmonies from a single melodic line. In my case I simply used slowly changing random V/Oct sampled values, and regularly trigger the Harmony module to reharmonize.

So without further ado, here is the patch and a sample video. You can get pre-release versions of Squinktronix and Venom plugins on GitHub. Or you can wait until the library releases.

The Venom Squinky Choir.vcv (4.4 KB)

And here is an old video I had never published using an older version of the patch as a backing track for a Native American flute solo. This patch version predates the development of the Venom plugin, so VCV Merge and Split modules were used to implement the recursion, just as Jakub originally demonstrated.


hey there, you are a bit alone with this subject, but what the hell, your patch is fantastic !

i don’t understand how it’s works, but i really enjoy the sound !!

i will take some time to understand what you do here, i love so much the taste of the sound !

thanks for sharing this patch and idea !

It isn’t choir like, but there are resonant filtered noise all over the 2016 Autechre live sets. At this point they perform with very complex Max/MSP parches, so you can pick out all sorts of basic synthesis: Kar plus-Strong, FM, filtered noise, etc.