The detectability of nocturnal carnivores is very low, because such data on many species can be difficult to acquire. Here, we design an automatic thermal video recording system for studying cryptic mammalian predators at food and water resources in Australia. We trialed different sampling procedures in contrasting environments between August 2005 and February 2006 to identify aspects of survey design that influenced video interpretability. We then compared the accuracy of the remote recording system to that of direct human observation. We found our system to be most effective when temperature settings were correctly adjusted and procedures were adapted for the sampling environment. We identified that the presence of vegetation and the distance between the target species and camera were the most important factors in video interpretability. The remote recording system was found to be as effective at recording both visits and behavior as direct human observation at a focal point. We propose that the automatic thermal video capture system has the potential to be an extremely useful tool for monitoring otherwise difficult to observe species. © 2013 The Wildlife Society.