This dataset contains the complete anonymized data used in the study that examined the acoustic and communicative characteristics of voices perceived as feminine, neutral, or masculine in Brazilian Portuguese. The dataset includes recordings from 47 Brazilian speakers of diverse gender identities (cisgender women, cisgender men, transgender women, transgender men, and non-binary individuals). Perceived gender was judged by 101 cisgender listeners using a 101-point visual analogue scale ranging from “very masculine” (−50) to “very feminine” (+50). Voices were categorized as feminine, neutral, or masculine using z-scores (±1 SD around the neutral midpoint). The dataset includes 16 acoustic measures extracted from sustained vowels, automatic speech, and expressive reading, using Praat scripts for jitter, shimmer, f₀ measures, HNR, H1–H2, spectral emphasis, CPPS, ABI, AVQI, speech rate, articulation rate, and prosodic variability. Additionally, 28 communicative parameters were evaluated by two experienced speech-language pathologists using the Brazilian protocol RoFOE (Roteiro Fonoaudiológico de Observação da Expressividade), covering aspects such as naturalness, confidence, pleasantness, resonance patterns, articulation precision, fluency, pause patterns, emphasis strategies, and verbal organization. All data were fully anonymized and contain no identifying information. This dataset supports the analyses reported in the manuscript and aims to promote transparency, reproducibility, and further research on gender perception, voice diversity, and communicative expressiveness in Brazilian Portuguese.