Virtual reality (VR) technology has become increasingly prevalent in telerehabilitation, due to its potential for providing immersive environments to enhance situation perception. Practical motion guidance optimizes motor training outcomes and facilitates skill learning and rehabilitation progress. Vibrotactile feedback has garnered attention due to the popularity of multimodal applications and has been shown to optimize motion guidance performance. However, the effectiveness and mental workload associated with visual and vibrotactile feedback instruction are currently unclear, and the research on the motion guidance mode is insufficient. Furthermore, the guidance performance (including direction accuracy, speed control accuracy, and completion time) and subjective mental workload of each guiding mode under static and dynamic movement states remain uncertain. To estimate the effectiveness of different feedback modalities, a 2 × 2 factorial within-subject design experiment involving 16 participants was conducted to assess the impact of feedback modalities and movement states on subjective mental workload. The results indicated a strong preference among participants for visual feedback, which was attributed to its higher motion accuracy and lower mental workload. Additionally, the dynamic movement state was associated with decreased motion guidance performance and increased mental workload. Future studies should explore performance under various motion-guiding scenarios to inform precise feedback mode design.