Determining regions in rendered sounds based on loudness contour (lots of FScape debugging as a "fun" side activity)

Some advances in developing the sound installation 'Through Segments' with a layer running on a Pi 4.

Definitely slower than on the laptop, but no real issues so far. Few xruns with binaural simulation and no jack params adjusted yet, so I'm fairly optimistic.

Have to run for long time to see if there are any memory leaks or such.

In 2011, I was pencilling around to understand how confluently persistent data structures work.

I'm adding a Histogram UGen to FScape; such a simple mechanism, just throw stuff into buckets and count… I will need this to look at the distribution of spectral features of a microphone signal.

Sometimes I wish I had more input on what building blocks other people would need, but I guess that's ok.

The plot shows a 17-bin histogram fed with a sine oscillator, and the histogram being flushed every ten steps.

Lastest Raspbian Buster comes with JDK 11, QJackCtl, SuperCollider 3.10.0 preinstalled; Mellite universal download was running out of the box 🎉

Seems much more powerful than the Pi 3, my bet is the 4 GB RAM are a significant point here with running the JDK app. So far all smooth ✔️

another, one of my favourites; I shot that landscape for the video piece 'Moor' (2016)

Show thread

Ok, I think I just accidentally created that Underworld "Rez" patch... [1/2]

(but Masto, why you cannot mix an image and a sound? see follow up toot)

Back to working with the genetic synth-def programming in ; I will try to integrate that soon into mainline Mellite. Now trying to automatically reduce the redundant graphs (replace constants, replace duplicate UGen signals). Here is a nice one:


(should I use mp3 directly now in mastodon or stick to funkwhale?)

Something weird happened to the high pass filter; I guess it didn't stay in sync with the frequency shifter?

Been hacking on a fully modulatable convolution UGen for (2). Seems to be working finally, allowing some funky things like quasi-continuous brickwall frequency filtering using dynamic windowed-sinc kernels 🤓 .

Next I want to try this as anti-aliasing filters for synchronous frequency shifting.

Making a postcard series for combo xCoAx 2020 / Almat 2020 conference and symposium (they will be available week after next at xCoAx 2019). I split the Almat motif set into two, using different parametrisation of the Neural Gas algorithm to trace my handwriting. I like both qualities, one is perhaps the gothic variant (gloomy like a Murnau), the other is more constructivist.

Status: rendering a series of images for an upcoming postcard. The sequence is - tablet/hand writing -> growing neural gas -> DSP compositing.

We're very happy with our three day workshop result - protoype Study (Nayarí Castillo, Emma Luke, Hanns Holger Rutz) :heart_cybre:

Show more

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!