Tonestacks
Pitch-aligned spectrum
Pair compatibility matrix
dark = smoother pair
Dyadic roughness
Harmony Field
Field idle
H --
C3-C6
Keys
Sampler
Fixed
Voices
4.8
2.0/s
Hamiltonian trace
H = --
Waiting for a generated state.
Model
\[
H(X)=\chi_B B(X)+\chi_P P(X)-\chi_R\lambda_R R(X)+\chi_S\lambda_I\bigl(I(X)-I^\star\bigr)^2
\]
\(X\) active notes · \(E_{ij}\) clipped 5-95% calibrated roughness · \(M\) decayed note memory · \(\mathcal{Y}\) previous sampled chords · \(\sigma=0.35\) · \(\kappa=3.0\)
pair field
--
\(B(X)=\sum_{\{i,j\}\subset X,\ i<j}E_{ij}\)
memory
--
\(P(X)=0.62\,\delta\left(
\frac{\sum_{x\in X}\sum_{m\ne x}M_mE_{xm}}{\sum_{x\in X}\sum_{m\ne x}M_m}
-0.26\min\left(1,\frac{\sum_{x\in X}M_x}{\sum_mM_m}\right)\right)\)
0.70
recurrence
--
\(R(X)=\max_{Y\in\mathcal{Y}}\exp\left[-D(X,Y)^2/(2\sigma^2)\right]\)
term \(-\lambda_RR(X)\)
0.35
surprisal
--
\(I(X)=-\log_2\frac{1}{|\mathcal{Y}|}\sum_{Y\in\mathcal{Y}}
\exp\left[-D(X,Y)^2/(2\sigma^2)\right]\)
term \(\lambda_I(I-I^\star)^2\)
1.6b
0.22
\(D(X,Y)=\frac12(d(X\to Y)+d(Y\to X))+0.18\frac{\bigl||X|-|Y|\bigr|}{\max(|X|,|Y|)}\)
\(d(A\to B)=\frac{1}{|A|}\sum_{a\in A}\min_{b\in B}E_{ab}\)
H --
I --
Sampler mode