When a Test Opens Another Path

Those tests helped me check how the machine worked, but they had a problem: they were too simple. Everything was too controlled and the movements too perfect for the situations I wanted to tackle when drawing more complex compositions.

Then I thought of reversing the approach behind KIKAIMA X SUMI. If I was already developing an interface to generate compositions from gestures, why not use that work to do the opposite? Rather than generating an image from scratch, I could start with a photograph, analyze its forms and turn them into paths the machine could draw.

That was how KIKAIMA X MICHI began.

The first results were quite good. Taking an arbitrary image, turning it into a line drawing and then seeing it on paper was easier than I expected. It also revealed limitations in the robot's construction, particularly in the mechanism that moves the tool towards the paper, which I had to modify.

But once those problems were solved, the tests began raising new questions. It was no longer just about reproducing the contours of an image. What if lines could vary in thickness? How could I interpret the contrast in a photograph to decide which areas needed greater presence and which could be resolved with lighter marks? What would happen if a line were built from other repeated lines?

What had initially been a tool for testing the machine began to become something else. Each new technical requirement opened a graphic possibility, and each possibility suggested more tests. Without planning it, I was developing a second system to explore different ways of turning images into drawings.

Although I still consider X MICHI a secondary branch of KIKAIMA, it has gradually acquired an identity of its own. I am especially interested in using my photographs and developing a way to interpret them that comes closer to how I draw or paint. I do not simply want the robot to copy an image, but to explore how it might translate it into a personal pictorial language.

In the end, that is what interests me most about MICHI's beginnings. I started by trying to check whether a machine worked properly and ended up wondering how it might learn to draw in a way I would recognize as my own.