Remnant is a multichannel acousmatic composition and live electronics system for orchestra and 8-channel diffusion. The project uses FluCoMa as a real-time analysis layer that modulates timbral tension continuously during performance.
The idea
The piece emerges from a biocentric reading of the Anthropocene: orchestral sounds treated as traces of a world that persists independently of human presence. The electronics do not accompany the orchestra — they listen to it, transform it, and relocate it within an artificial sonic environment that does not document but hypothesises.
The compositional system navigates a timbral space built from the ConTimbre corpus (orchestral multiphonics, extended techniques, spectral sounds) through Brownian motion, organised according to McAdams’ timbral prolongational hierarchy and Thoresen’s Dynamic Forms (Aural Sonology).
Theoretical framework
McAdams’ timbral prolongational hierarchy treats timbre as a structural force analogous to harmony in tonal music. Sounds are positioned in a perceptual space defined by spectral and temporal features (brightness, roughness, attack time, spectral flux); proximity in that space creates timbral coherence, distance creates tension. The hierarchy describes how timbral events prolong, destabilise and resolve — providing a grammar for non-pitched, spectrally-defined music.
Lerdahl’s timbral tension extends this into a quantitative model: tension is a function of distance within the timbral space, weighted by the hierarchical position of each event. In Remnant, this value governs the convolution engine — at low tension the orchestral source remains recognisable; at high tension it is progressively masked and displaced into an artificial acoustic environment.
The two models work at different scales: McAdams structures the compositional trajectory (which sounds appear, in what order, across how long); Lerdahl governs the real-time processing (how those sounds are transformed moment to moment).
FluCoMa in the system
FluCoMa handles the real-time analysis layer in `remnant_flucos.scd`. The orchestra signal (4 field tracks via BlackHole → `SoundIn`) feeds:
- `FluidOnsetSlice` — onset detection driving gestural punctuation
- `FluidMFCC` + `FluidStats` — timbral feature extraction feeding a tension estimator
- `FluidNoveltySlice` — structural boundary detection
The output is a continuous tension value (`~r_flucoTension`) that modulates the convolution processing with a maximum contribution of 0.4 on top of the score-defined tension. Where the FluCoMa analysis and the compositional trajectory converge, the field is coherent; where they diverge, stochastic geometry and live material pull in opposite directions.
The composite tension is: `finalTension = (scoreTension + flucoTension × 0.4).clip(0, 1)`
This value is visible in real time on the HUD as a colour-coded bar alongside the score tension, so the performer can read both streams simultaneously.
Architecture
```
Orchestra → BlackHole → SoundIn (4 fields)
→ FluCoMa analysis → \~r_flucoTension
→ Lerdahl masking pipeline (per field):
PitchShift → Convolution2 → FreqShift → BPF
→ 8-channel octophonic output
```
IRs are generated procedurally in SC — one per Lachenmann timbral category (Farbklang, Geräusch, Klang, Kadenz, Textur, Stille, Neutro). All fields share the same IR and tension value, treating the orchestra as a single sonic body.
Performance
In concert, gesture categories are triggered manually by the performer via seven buttons on the HUD (one per Lachenmann category). The performer listens to the orchestra and decides when to switch — there is no automatic timer. Timbral tension is controlled in real time via a slider or NanoKONTROL2 (CC 0 / knob 1).
FluCoMa analysis runs continuously in the background, adding automatic modulation on top of the performer’s tension setting. The tension estimator currently works but could be more sophisticated — any thoughts on more robust real-time timbral feature extraction for this kind of continuous modulation would be very welcome.
Repository