The dataset consists of simultaneous recordings of HD EMG signals from proximal muscles of the forearm and hand kinematics estimated from five cameras. The myoelectric activity of the flexor and extensor finger muscles was recorded using an HD EMG system with two grids of 64 electrodes positioned on the participant’s forearm. Signals were recorded in monopolar mode at a sampling rate of 2052 Hz. At the same time, five high-speed cameras (DFK-37BUX287, The Imaging Source) placed around a cubic structure (0.34 m³) recorded hand videos at 100 FPS with a resolution of 720 × 540 pixels. MediaPipe and Anipose were used to reconstruct the 3D kinematics of finger movements captured by the five cameras. The cameras and EMG system were synchronized by a synchronization station that generated a signal from a microcontroller. Participants performed manual dexterity tasks in a seated position. The tasks included two different grips (one- and two-finger pinch) and a full grasping movement, performed under 30 conditions combining five wrist angles (0° to 180° in steps of 45º) and two elbow extensions (flexed and extended). The execution order was randomized, with each task lasting 30 s. A virtual hand model was also displayed on a monitor during the experiments, showing the movement frequency (0.50 Hz) that participants had to follow. In total, twenty-three healthy individuals, divided into groups of ten across age ranges from 20 to 70 years, voluntarily took part in the experiments and signed informed consent. The protocol was approved by the Ethics Committee of the University of Campinas (CAAE: 80299124.2.1001.5404). This dataset is part of the MYOREHAB consortium supported by FAPESP (EU-LAC Joint Call FAPESP - Proc. no. 2022/05104-8).