With the change of net primary productivity in a natural forest of Castanopsis carlesii in subtropical region under climate change, litter and plant root inputs will also change, which will significantly affect soil microbial community. In this study, the Detritus Input and Removal Treatments (DIRT) had been set up in a natural C. carlesii forest for seven years. In July 2019, soils of two layers (0—10, 10—20 cm) indifferent treatments (control, removal of above-ground litters, removal of roots, no input of litters, and addition of double above-ground litters) were collected to measure microbial phospholipid fatty acid (PLFA) contents. The ratio and diversity of microbial groups were calculated to further reveal the influencing mechanism of litter and plant root inputs on soil microbial community composition and diversity in a subtropical natural C. carlesii forest. The results showed that: (1) The PLFA content of microorganisms in the 0—10 cm soil layer for different treatments was about twice that in the 10—20 cm soil layer. (2) The addition or removal of litters reduced the contents of gram-positive bacteria, gram-negative bacteria, actinomycetes and other bacteria to varying degrees, but had no effect on arbuscular mycorrhizal fungi. The root removal reduced the PLFA content of arbuscular mycorrhizal fungi significantly. (3) Shannon-Wiener and Simpson diversity indices of the microbial communities were not affected by different treatments. Litter removal reduced the richness of soil microbial community and improved the evenness. The content and distribution of microbial community in 0—10 cm soil layer were greatly affected by the change of litter input. (4) Plant roots improved the content of fungi, while litter input mainly changed bacterial abundance and structure. Soluble organic carbon and mineral nitrogen were the main factors that affect the microbial community composition and diversity. There fore, litter and root inputs affected soil microbial community through the change of soil properties. The results can provide a scientific basis for understanding the effects of interactions among plants, soil, and microorganisms on forest productivity.