This paper addresses bimanual robot manipulation tasks where coordinated actions between two robotic arms are required for complex object interactions. The work introduces a Collaborative Preparatory Manipulation framework that enables robots to perform sequential preparatory actions - such as pushing objects to accessible positions or lifting items - to facilitate subsequent goal-directed manipulations by the other arm. The proposed visual affordance-based approach first anticipates the final task objective and then generates appropriate preparatory manipulations, requiring deep understanding of object geometry, spatial relationships, and semantic properties. By learning from demonstrations and employing vision-based affordance recognition, the framework achieves effective bimanual coordination for tasks involving objects that are difficult to grasp directly.