Sampling animal populations with camera traps has become increasingly popular over the past two decades, particularly for species that are cryptic, elusive, exist at low densities or range over large areas. The results have been widely used to estimate population size and density. We analyzed data from 13 camera trap surveys conducted at five sites across the Kaa-Iya landscape, Bolivian Chaco, for jaguar, puma, ocelot and lowland tapir. We compared two spatially explicit capture–recapture (SCR) software packages: secr, a likelihood-based approach, and SPACECAP, a Bayesian approach, both of which are implemented within the R environment and can be used to estimate animal density from photographic records of individual animals that simultaneously employ spatial information about the capture location relative to the sample location. As a non-spatial analysis, we used the program CAPTURE 2 to estimate abundance from the capture–recapture records of individuals identified through camera trap photos combined with an ad hoc estimation of the effective survey area to estimate density. SCR methods estimated jaguar population densities from 0.31 to 1.82 individuals per 100 km2 across the Kaa-Iya sites; puma from 0.36 to 7.99; ocelot from 1.67 to 51.7; and tapir from 7.38 to 42.9. Density estimates using either secr or SPACECAP were generally lower than the estimates generated using the non-spatial method for all surveys and species; and density estimates using SPACECAP were generally lower than that using secr. We recommend using either secr or SPACECAP because the spatially explicit methods are not biased by an informal estimation of an effective survey area. Although SPACECAP and secr are less sensitive than non-spatial methods to the size of the grid used for sampling, we recommend grid sizes several times larger than the average home range (known or estimated) of the target species.