Dynamic hand gesture recognition (HGR) is critical for real-time prosthetics and neural robotics applications, where high-density surface electromyography (HD-sEMG) signals provide valuable insights into muscle activity. However, adapting machine learning models to new users while main-taining high accuracy and computational efficiency presents a significant challenge, mainly when dealing with individual variability in muscle signals. This study investigates the application of Low-Rank Adaptation (LoRA) in transformer-based models for dynamic HGR tasks. By incorporating LoRA into a pre-trained generalized transformer model, we explore its effectiveness in adapting to new subjects with varying rank values (r = 32, 64, 96) and different HD-sEMG signal window sizes (100 ms, 200 ms), Our experimental results demonstrate that LoRA achieves high accuracy during the dynamic phase of hand gestures and improves computational efficiency by reducing training parameter size. It is a promising solution for scalable and robust transformer-based systems in dynamic HGR. Furthermore, the ability to adapt quickly to new subjects with minimal retraining could make this approach particularly valuable for real-world applications.