At their home in rural China last year, my grandparents found a small piece of old paper tucked away in a book. It was a love letter that my dad wrote my mom in the '90s when they were long-distance dating, complete with frayed, folded edges and smudged blue ink. When I saw it I was immediately emotional, nostalgic for a time when loving was hard, but worthwhile. Letters to My Mom is a bittersweet homage to my parents' story and a machine exploration of loss, labor, and love.

I cropped, annotated, and upscaled each character of my dad's original letter captured by my grandpa's phone camera halfway across the world. Using this small dataset, I trained a generative Chinese character model that imitated the ink, stroke, and paper my dad used. I then typed up two love letters to my mom and to my partner, fed these characters through the model I trained to generate 256px square images of each unique character, and printed out each resulting "handwritten" letter on sheets of canvas. One process of labor replaces another. When the viewer looks carefully, they see smudges and imperfections in the blue ink, but the naturality also feels eerily uniform and rigid. From the human past to the machine future, is there always something that is lost?

Jian Zhang's letter to Yinghui Gu, 1999

Letters to my Mom, 2025