Fifty-six pattern works from the project, drawn from all three combination methods across the five patterns. Method 1 is multi-target training: a single network learns both patterns and interpolates internally. Method 2 is fine-tuning interpolation: a model trained on one pattern is fine-tuned toward another, and the two weight sets are blended. Method 3 is weight interpolation between independently trained models. Each work is a single frame from an evaluation sweep, realised through a palette pipeline of gradient mapping, optional inversion, gamma, and grain.

In each work’s label, m = 0 corresponds to the first parent listed and m = 1 to the second.