Conventional VR mining safety training relies on handheld controllers that lack authentic gesture replication, compromising procedural accuracy and emergency skill transfer, thereby increasing real-world operational risks. This study resolves the interaction authenticity limitations of handheld controllers in VR mining training by developing a gesture-driven system using Meta Quest3’s markerless tracking. The system combines artificial intelligence generated mining scenes with specific task gestures such as fire extinguisher operation and valve rotation through Unity3D. A 30-participant experiment using System Usability Scale and Igroup Presence Questionnaire metrics showed the gesture group achieved 34% higher usability and 12% greater immersion, with 90% endorsing superior realism. Results confirm controller-free gestures enhance physical mapping and cognitive engagement in industrial VR training.