This work presents a new approach to enable accurate machining cycle-time prediction. While conventional computer-aided manufacturing (CAM) software estimates cycle times for given toolpaths, huge errors occur due to neglecting interpolator dynamics of numerical control (NC) systems. Typically, the machine changes its feedrate such as slowing down and speeding up depending on the existence of junction points between consecutive waypoints to realize smooth motion without violating unacceptable trajectory error. In order to capture and utilize such machine-specific dynamic behavior, a cycle-time prediction framework is proposed in combination with an analytical jerk-limited feedrate planner (JFLP) and data-driven artificial neural networks (ANNs). The high-level interpolation models are first defined as local corner interpolator and global corner interpolator. Then, the cycle-time prediction framework is presented that analyzes geometric characteristics of given toolpaths, classifies types of interpola- tion models for every corner, and predicts parametric information by ANNs necessary for JLFP to plan kinematic profiles. In addition, the process to extract ground truth information, which is utilized for training ANNs, based on toolpath geometry and pre-recorded kinematic profiles is explained in detail. In order to validate its performance, several experiments including data collection for training are conducted using the actual CNC machine. Experimental results against realistic toolpaths validate its effectiveness in predicting cycle times accurately.