PMA (propidium monoazide) is one of the few methods that are compatible with metagenomic sequencing to characterize the live/intact microbiota. However, its efficiency in complex communities such as saliva and feces is still controversial. An effective method for depleting host and dead bacterial DNA in human microbiome samples is lacking. Here, we systematically evaluate the efficiency of osmotic lysis and PMAxx treatment (lyPMAxx) in characterizing the viable microbiome with four live/dead Gram+/Gram- microbial strains in simple synthetic and spiked-in complex communities. We show that lyPMAxx-quantitative PCR (qPCR)/sequencing eliminated more than 95% of the host and heat-killed microbial DNA and had a much smaller effect on the live microbes in both simple mock and spiked-in complex communities. The overall microbial load and the alpha diversity of the salivary and fecal microbiome were decreased by lyPMAxx, and the relative abundances of the microbes were changed. The relative abundances of Actinobacteria, Fusobacteria, and Firmicutes in saliva were decreased by lyPMAxx, as was that of Firmicutes in feces. We also found that the frequently used sample storage method, freezing with glycerol, killed or injured 65% and 94% of the living microbial cells in saliva and feces, respectively, with the Proteobacteria phylum affected most in saliva and the Bacteroidetes and Firmicutes phyla affected most in feces. By comparing the absolute abundance variation of the shared species among different sample types and individuals, we found that sample habitat and personal differences affected the response of microbial species to lyPMAxx and freezing. IMPORTANCE The functions and phenotypes of microbial communities are largely defined by viable microbes. Through advanced nucleic acid sequencing technologies and downstream bioinformatic analyses, we gained an insight into the high-resolution microbial community composition of human saliva and feces, yet we know very little about whether such community DNA sequences represent viable microbes. PMA-qPCR was used to characterize the viable microbes in previous studies. However, its efficiency in complex communities such as saliva and feces is still controversial. By spiking-in four live/dead Gram+/Gram- bacterial strains, we demonstrate that lyPMAxx can effectively discriminate between live and dead microbes in the simple synthetic community and complex human microbial communities (saliva and feces). In addition, freezing storage was found to kill or injure the microbes in saliva and feces significantly, as measured with lyPMAxx-qPCR/sequencing. This method has a promising prospect in the viable/intact microbiota detection of complex human microbial communities.