未知 Michi — Courses

Great images rarely arrive fully formed. Gerhard Richter paints, photographs, then paints again; Agnes Martin drew grids, erased, and redrew until the tremor was right. Multi-stage refinement means generating an initial image, diagnosing its specific weaknesses (color drift, compositional imbalance, lost detail), then regenerating with targeted prompt corrections. The discipline is analytical: each iteration must fix at least one named flaw without introducing new ones.

Monet painted 30 Rouen Cathedrals not from indecision but from systematic investigation — same subject, varied conditions. Warhol's repetition-with-difference revealed how small changes (color, exposure, registration) alter meaning entirely. Controlled variation means generating multiple versions of a single concept where ONE parameter changes while everything else holds constant. This isolates variables and builds understanding of what each prompt element actually controls.

The Oulipo writers chose arbitrary constraints (no letter 'e' in an entire novel) and found that limitations liberated creativity. Lars von Trier's Dogme 95 banned tripods, artificial lighting, and post-production — and produced some of cinema's rawest emotional power. Sol LeWitt wrote instructions so strict that anyone could execute them, yet the results were undeniably art. Productive constraints mean choosing extreme limitations (single color, maximum 3 shapes, fixed aspect ratio, banned subjects) and discovering that restriction forces invention.

Kandinsky heard colors and painted sounds. Paul Klee structured canvases like fugues — theme, variation, inversion, recapitulation. Visual music means translating temporal structures into spatial ones: tempo becomes density of marks (fast staccato brushwork vs slow gradient washes), crescendo becomes escalating scale or saturation, silence becomes negative space functioning as rest, and counterpoint becomes two visual themes coexisting in productive tension. The image should feel like it has rhythm even though it is still.

Cinema developed a visual grammar over 130 years: close-ups create intimacy, wide shots establish context, low angles convey power, high angles suggest vulnerability. Kubrick composed frames with obsessive symmetry; Tarkovsky held shots until time itself became palpable; Wong Kar-wai's color grading turned Hong Kong into a neon fever dream. Cinematic thinking means treating each image as a frame from an unmade film — with implied narrative before and after, deliberate camera placement, and mise-en-scène where every element in the frame serves the story.

Architecture is the art of structured emptiness — the space between walls matters as much as the walls themselves. Tadao Ando's concrete creates spiritual silence; Zaha Hadid's fluid forms defy gravity and convention; Luis Barragán's color walls transform light into emotion. Architectural thinking in image-making means using structure as metaphor (a column is authority, an arch is transition), negative space as inhabitable room (emptiness the viewer can psychologically enter), scale as emotional instrument (a tiny figure against a vast wall is sublime), and material weight as gravity (heavy stone grounds, glass liberates).

Dance is the art of the body in space and time. Degas spent decades capturing dancers in moments between poses — the unguarded instant when a body is most truthful. Muybridge's motion studies decomposed galloping horses into frozen instants and rewrote how we understand movement. The Futurists (Boccioni, Balla) painted speed itself — lines of force, blurred repetition, the trajectory of a body through time. Frozen motion in a still image means capturing the exact moment that implies both where the body came from and where it is going. Gesture becomes communication; the angle of a hand, the arc of a leap, the twist of a torso all carry meaning.

Every medium has edges — things it can do brilliantly and things it cannot do at all. Photography cannot show thoughts; sculpture cannot show time; painting cannot show sound (though Kandinsky tried). AI image generation has its own frontier: it excels at texture, atmosphere, and impossible combinations, but struggles with precise counting, spatial consistency, text rendering, and logical relationships. Mapping these frontiers is not failure analysis — it is research. Documenting what breaks, how it breaks, and what emerges from the breaking is the first step toward original work.