Large carnivores play a pivotal regulating role in maintaining healthy and balanced ecosystems; however, most of them are rare and elusive, and knowledge about their resource consumption is scarce. Traditional methods based on morphological identification of undigested remains are labor intensive and often not sufficiently accurate, leading to errors and biased ecological inferences. Here, we developed a multi-marker DNA metabarcoding approach to analyze the dietary diversity of giant otters (Pteronura brasiliensis) from fecal DNA while controlling predator species identity. We combined two mitochondrial markers, 12S rRNA and cytochrome c oxidase 1 (COI) gene, that target the full range of potential vertebrate and invertebrate prey. We compiled a local reference database of DNA barcodes for most potentially ingested fish, which were used to evaluate the specificity of the metabarcoding primers in silico. Most prey are identified at the species level (>90%) and the dietary profiles provided independently by the two markers are highly similar, whether in terms of list of prey or frequency of occurrences, hence validating the approach. We detected a higher number of rare fish prey with the 12S primers that amplified solely Teleost species while the degenerate COI primers revealed non-fish prey (e.g., amphibians, snakes, birds, and earthworms) and confirmed predator species identity. This study demonstrated that scat DNA metabarcoding is particularly useful to provide in-depth information on elusive carnivorous dietary profile. Our methodology opens up new opportunities to understand how top carnivores diet cope with the effects of anthropogenic alteration of ecosystems and identify conflicts with humans and livestock.