UNSW Sydney researchers have developed a breakthrough deep learning model capable of detecting blue whale songs using only a single audio recording. Published in Scientific Reports, the study led by Ph.D. candidate Ben Jancovich demonstrates that this approach can identify whale calls in vast acoustic archives spanning decades and entire ocean basins. Traditionally, training such models required thousands of labeled recordings, a significant hurdle for rare or elusive species where data is scarce. The team overcame this limitation by leveraging a specific characteristic of blue whales: their highly stereotyped vocalizations. Unlike dolphins, whose individual whistles vary greatly, blue whales within the same population produce nearly identical calls. This consistency allowed researchers to generate a robust training dataset from just one original example. By applying data augmentation techniques such as pitch shifting, time stretching, and adding background ocean noise, they created thousands of semi-synthetic recordings that simulated natural variations and sound propagation conditions. Tests on real-world data confirmed the method's efficacy. When applied to recordings of pygmy blue whales, the detector correctly identified 99.4% of the calls, performing on par with models trained on much larger datasets. The innovation also offers significant environmental and computational benefits. Standard deep learning training often demands immense computing power and electricity, but this streamlined model can be trained on a standard laptop in hours rather than weeks. This efficiency makes the tool accessible to researchers without access to high-performance supercomputers. The implications for ecological research are profound. Vast amounts of underwater data have been collected globally through passive acoustic monitoring, yet much remains underutilized because manual analysis is impossible for long-term datasets and automated tools lack training data. This new system promises to unlock these archives, enabling scientists to track long-term changes in marine mammal populations and behavior. The research team plans to apply the detector to a 25-year dataset from the central Indian Ocean to monitor shifts in blue whale song over time. Beyond population monitoring, the technology could provide insights into animal culture, such as how songs are learned and transmitted across generations. Furthermore, the method is not limited to marine life; it could be adapted to monitor other species with consistent vocal patterns, including birds and insects, in diverse environments like forests. By enabling the study of rare species from a single recording, this tool represents a major step forward in conservation and the understanding of animal communication.