This chapter explores solutions for designing bandwidth-adaptive semantic communication systems that can operate efficiently under varying bandwidth constraints. It addresses the limitations of using multiple fixed-bandwidth models, which introduce significant storage overhead and latency. The chapter presents two key adaptive strategies. Entropy-aware methods, inspired by traditional coding, allocate more symbols to features with higher information density, improving performance by focusing on important data. Semantic-aware methods, a learning-based approach, dynamically assign bandwidth based on the semantic importance of content for a specific task. These strategies enable a single system to achieve performance comparable to specialized models while drastically reducing overhead, paving the way for more practical, real-world applications.