Binding is key in multisensory perception. This study investigated the audio-visual (A-V) temporal binding window in 4-, 5-, and 6-year-old children (total N = 120). Children watched a person uttering a syllable whose auditory and visual components were either temporally synchronized or desynchronized by 366, 500, or 666 ms. They were asked whether the voice and face went together (Experiment 1) or whether the desynchronized videos differed from the synchronized one (Experiment 2). Four-year-olds detected the 666-ms asynchrony, 5-year-olds detected the 666- and 500-ms asynchrony, and 6-year-olds detected all asynchronies. These results show that the A-V temporal binding window narrows slowly during early childhood and that it is still wider at 6 years of age than in older children and adults.