AttractorScope

Hi everyone,

I have been working recently on the AttractorScope class which visualizes audio signals as phase-space attractors in real-time, from 2D up to 6D.

It uses delay coordinate embedding (Takens’ theorem) to reconstruct the underlying dynamics of any audio or control bus signal. Essentially, it takes time-delayed copies of a signal and plots them as coordinates in N-dimensional space, revealing the attractor geometry of the sound.

Features include:

  • Multiple rendering styles (lines, points, glow, ribbons, heat maps)
  • Dynamic color modes (rainbow, velocity-based, distance-based, curvature-based)
  • Interactive rotation and real-time parameter control
  • Works with any audio or control bus

I made it primarily for aesthetic visualization, but it could potentially be useful for research.

Installation: Available as a Quark - just run:
Quarks.install("https://github.com/Kosmas-Giannoutakis/AttractorScope.git")

I’d love to hear feedback, suggestions, and contributions from the community! Bug reports, ideas, and pull requests are all welcome.

GitHub: GitHub - Kosmas-Giannoutakis/AttractorScope

Best,
Kosmas

8 Likes

Thank you for this — it’s amazing!

I was wondering: is it possible to display both the two‑sided and the one‑sided phase, as illustrated here, simply by adjusting parameters? It seems impossible to me, but I am asking in case there is a possibility.

I may be wrong, but I think the phase space analysis from dynamical systems is unrelated to the phase information in fft analysis.

However, it could be worth experimenting with applying delay coordinate embeddings to the frequency domain of audio signals.

1 Like