The ability to produce light has convergently evolved dozens of times on this planet. However, given the diversity of luciferases and luciferins, it nonetheless remains a challenge to find a common principle for how a luciferase could evolve from a non-luminous protein. Gene duplications are responsible for many crucial changes over evolutionary time, from color vision in primates to bone formation in vertebrates. The principle is that a duplicated gene is free to evolve a new function, a process called neofunctionalization. Previous research has identified that many luciferases are related to other enzymes not involved in luminescence, thus are likely cases of neofunctionalization. Here I examine the role of gene duplications on the origin of bioluminescence in animals, considering cases in cnidarians, arthropods, and molluscs. Such duplications are widespread, found in most cases where a luciferase or photoprotein has been cloned, in some cases emerging from very large protein families. Octocorals and squid appear to have a single specialized enzyme emerging in a few species following a duplication affecting the whole clade. However, in the adenylating enzymes, fireflies, squid, and luminous flies all have large lineage-specific expansions, of which only a few members are identified as luciferases. Rather unusually, a few luciferases (Vargula luciferase and Gaussia luciferase), appear as sequence isolates, with no similar enzymes in public databases. Given the parallel origins of bioluminescence in certain protein families, particular proteins may have a propensity for evolving this function, perhaps by having affinity for lipophilic molecules whereupon a single point mutation allows the coordination of molecular oxygen.