Generative art series: make variations worth keeping

A series has selection criteria

A series asks more of a generator than a single striking result. Every image needs a family resemblance, but the set also needs meaningful differences. Begin by naming the qualities you want to preserve. Those become editing criteria rather than accidental habits of the current code.

Expose three controls and record each output

Choose one existing sketch and expose no more than three artistic controls. Give each control a bounded range and save a record of the exact values used for every output. Include the seed when randomness matters. Render a small overview before committing to larger files.

Set up the first proof around exclusion area, using a value you can recognize in the finished result. Begin with the smallest version that demonstrates the proposed relationship. Remove secondary decoration while checking its construction. A tiny proof makes incorrect assumptions visible before they multiply across the finished work. A useful acceptance check is whether the sequence changes its spatial rhythm while retaining a shared system. Judge that at the intended viewing size or working distance, where tiny construction mistakes and exaggerated effects can have very different consequences.

Distinguish structural change from detail

If every variation feels interchangeable, your controls may change details while leaving the dominant structure untouched. If the set feels unrelated, too many controls may vary simultaneously. Freeze the strongest structural control and examine one axis of variation at a time.

Before rebuilding everything, compare the current exclusion area setting with the last understandable version. Temporarily show construction points and intermediate values. Reduce repeated elements to a handful and pause anything that changes over time.

Edit five images into a sequence

Select five images to form a sequence with a beginning, a transition, and a destination. Remove the most spectacular one if it breaks the sequence. Explain what the remaining set gains through that omission.

Reserve one trial each for small, intermediate, and nearly half the field. Place the results together, or replay them in a fixed order when movement matters. Look for the smallest change that alters the reading of the work. If every trial produces a different problem, reduce the distance between the settings and investigate the transition instead of choosing an extreme.

Name files for reconstruction

For the opening-field series, use a filename that records the seed and exclusion size, such as field-seed17-opening80. Keep dimensions and any other fixed controls in a small accompanying note. The filename need not contain the whole recipe, but it should lead back to it. Save unselected proofs separately from the final sequence so a later revision can compare alternatives without confusing them with the intended reading order.