Passive acoustic monitoring promotes a deeper understanding of the biodiversity and health of marine ecosystems, with potential to fulfill scientific, policy, and industry needs. A large bottleneck has been the human effort required to find and label diverse and abundant fish call types in acoustic recordings. Training machine learning models on different recording devices is paramount to accelerating soundscape characterizations. Here, we labeled fish calls to improve machine learning models on automatically detecting fish calls across  five tropical coral reefs in the U.S. Virgin Islands and Hawaii. Such models have potential to be rapidly scaled for continuous ecosystem monitoring, improving estimates of fish and invertebrate density and distribution, and inform acoustic enrichment for reef conservation and restoration. Collaborator