During opencast mining of hard rock deposits, maximizing explosive energy efficiency improves fragmentation and rock displacement. High-speed videography enables millisecond-scale analysis of blast dynamics, facilitating the assessment of how explosive energy distribution influences burden movement and velocity, both of which are critical for fragmentation efficiency, safety, and overall blast performance. Determining the burden rock velocity requires high-speed videography and advanced processing software, rendering it a costly and time-intensive process. In this context, a study is proposed to predict Burden Rock Velocity (BRV) in limestone bench blasting using high-speed videography. Field visits were conducted at three mechanized limestone mines in southern India, where data from 166 blasts were collected. BRV was determined using high-speed videography and analyzed with ProAnalyst software. Eight key input parameters such as bench height (BH), total explosive charge (TEC), stiffness ratio (K), charge factor (CF), joint spacing (Js), uniaxial compressive strength (UCS), P-wave velocity (P-W), and rock density (ρ) were used to develop a predictive model for BRV based on a hybrid extreme gradient boosting (XGBoost) algorithm optimized with Ant Colony Optimization (ACO). The hybrid XGBoost–ACO model was compared with the baseline XGBoost model. Results demonstrated that the optimized XGBoost–ACO model achieved superior performance, with an R2 of 0.964, RMSE of 0.14, and MAE of 0.272, outperforming the baseline XGBoost model (R2 = 0.909).