Kimberly Darwin

I was making paintings move before AI learned how — hand-animating a five-hundred-year-old portrait frame by frame because I wanted him to say something ridiculous. The tools finally caught up to the impulse. What I've learned from these tools: the motion must come from inside the piece. I don't animate paintings; I ask what a painting is already doing and let it keep going. A meadow of heart-shaped flowers wanted wind. Clouds want to wisp by. When the concept comes from outside the work, it sits on the surface like a filter. When it comes from inside, the painting seems to have always been moving.

The tools are instruments, not answers — so I test them like an art director, not a fan. I ran the line drawing below through four different models: most rendered beautiful glass, but only one understood that a hole goes all the way through. Fidelity isn't a function of price; you only learn which model can hear you by asking all of them. I match platform to problem — hosted models for speed, node-based pipelines when I need control, hand animation only when the machines can't hear what I'm asking — and my first take is always a hypothesis, not a result. Sometimes the model's mistake is better than my plan; it’s a melding of minds, after all. AI hasn't changed what I believe about making things. It still comes from within, even if we get some help articulating our ideas.