The unit-dose repackaging of medications involves cutting blister packs to separate each pill while maintaining the integrity of the original packaging. Subsequently, the pills are repackaged, including essential information for medication identification. The medication-cutting process is currently carried out manually in most hospitals, and automating it would make this process more efficient and optimized. This paper aims to generate possible cutting paths to unit-dos repack blister packs automatically blister packs. The proposed approach uses a convolutional neural network (Yolov8) to detect the pills, combined with the Voronoi Diagram algorithm to generate the cutting paths. As a result, the detection network trained with 8,856 labeled images of blister-pack and pills achieved a 99.0% mean average precision (mAP). Moreover, 60 blister layouts were used, and the proposed algorithm successfully predicted and generated cutting paths for 48 types.