High-performing machine translation (MT) systems can help overcome language\nbarriers while making it possible for everyone to communicate and use language\ntechnologies in the language of their choice. However, such systems require\nlarge amounts of parallel sentences for training, and translators can be\ndifficult to find and expensive. Here, we present a data collection strategy\nfor MT which, in contrast, is cheap and simple, as it does not require\nbilingual speakers. Based on the insight that humans pay specific attention to\nmovements, we use graphics interchange formats (GIFs) as a pivot to collect\nparallel sentences from monolingual annotators. We use our strategy to collect\ndata in Hindi, Tamil and English. As a baseline, we also collect data using\nimages as a pivot. We perform an intrinsic evaluation by manually evaluating a\nsubset of the sentence pairs and an extrinsic evaluation by finetuning mBART on\nthe collected data. We find that sentences collected via GIFs are indeed of\nhigher quality.\n