The curator's reviewBreed images like show dogs. Cross a portrait with a nebula, nudge the genes for age or chaos, and watch offspring inherit the strangest features of both parents.
Artbreeder began life in 2018 as Ganbreeder, Joel Simon's experiment in letting people wander a GAN's latent space together, and the central metaphor stuck: images have genes. Take a portrait, cross it with a nebula, and the child inherits from both parents. Nudge sliders labeled things like age, chaos, or fur, and you are doing husbandry on the space of possible pictures.
The mechanics reward patience. Every image on the site can be bred with every other; children can be bred back with parents; a slider drag is a mutation. People ran deliberate multi-generation breeding programs to reach a face or a landscape they could see in their heads, and some bloodlines on the site are genuinely generations deep, with family trees you can walk backward like ancestry records.
Its cultural moment was enormous and is now half-forgotten: for a stretch of 2019 to 2021, before diffusion models made images cheap, Artbreeder portraits were everywhere - character art for novels, faces for tabletop campaigns, entire fandoms breeding their protagonists. It was many people's first hands-on experience of latent space, and the interface taught the concept better than any diagram: similar images live near each other, and you can walk between them.
The vocabulary is what stays with me: genes, children, lineage, crossbreeding - selective breeding applied to the dream-space of a network. Husbandry for pictures. The herd is still there, still breeding, and the family trees are still growing.
