Word frequency, context, and length are three core elements that impact speech perception. Considering the limitations of previous Chinese stimulus databases, such as non-standardized sentence structures, uncontrolled emotional information that may exist in semantics, and a relatively small number of voice items, we developed an abundant and reliable Chinese Mandarin nonsense pseudo-sentences database with fixed syntax (pronoun + subject + adverbial + predicate + pronoun + object), lengths (6 two-character words), and high-frequency words in daily life. The high-frequency keywords (subject, predicate, and object) were extracted from China Daily. Ten native Chinese participants (five women and five men) evaluated the sentences. After removing sentences with potential emotional and semantic content valence, 3,148 meaningless neutral sentence text remained. The sentences were recorded by six native speakers (three males and three females) with broadcasting experience in a neutral tone. After examining and standardizing all the voices, 18,820 audio files were included in the corpus (https://osf.io/ra3gm/?view_only=98c3b6f1ee7747d3b3bcd60313cf395f). For each speaker, 12 acoustic parameters (duration, F0 mean, F0 standard deviation, F0 minimum, F0 maximum, harmonics-to-noise ratio, jitter, shimmer, in-tensity, root-mean-square amplitude, spectral center of gravity, and spectral spread) were retrieved, and there were significant gender differences in the acoustic features (all p < 0.001). This database could be valuable for researchers and clinicians to investigate rich topics, such as children’s reading ability, speech recognition abilities in different populations, and oral cues for orofacial movement training in stutterers.