To overcome limitations of flexible thermoacoustic devices in low-frequency emission and sensing, we present a resonance-adjustable graphene sound device (RAGSD) inspired by frog vocal sacs. Integrating laser-induced graphene (LIG) with a deformable cavity, RAGSD provides continuous tuning from 922.12 to 1762.90 hertz. A dynamic, continuously tunable electro-mechano-acoustical model explains the mechanism and predicts frequency with
R2=0.990
. In emission, a 25.34-decibel sound pressure level gain was observed at resonance, enabling controllable, frequency-selective voice amplification for personalized output. In sensing, the LIG piezoresistive readout delivers sensitive transduction, while resonance matching amplifies weak, high-frequency cardiac sounds. Wearable tests on healthy volunteers and patients recorded clear S1/S2 and pathological murmurs. Integrated with AuscNet-H, a deep learning algorithm designed for heart sound classification, the system achieved 99.375% accuracy across four clinical classes. Under inflation, no false negatives occurred, and misclassifications among similar diseases were fewer than with a commercial electronic stethoscope. These results demonstrate a practical path toward intelligent, wearable auscultation.