Processing speech in noisy environments is a core challenge for children with attention deficit hyperactivity disorder (ADHD), and few studies investigated evidence-based interventions. Disrupted visual attention and audiovisual integration are key contributors to these difficulties. Virtual reality (VR) games offer potential to support attention and audiovisual integration training, yet few are designed for this purpose. Current approaches using VR games emphasize either clinical fidelity or entertainment, creating an imbalance between neural engagement and sustained motivation. We developed a VR game through a co-design approach, informed by neural mechanisms and clinical expertise, and embedding child-centered interaction to sustain engagement. The game includes variations in interaction modes and difficulty levels, iteratively refined with neurodevelopmental specialists and children. We conducted an exploratory study with 11 participants, including neurotypical and ADHD children. Findings highlight task performance and insights from the target user group, while also suggesting implications for balancing clinical potential with user engagement.