In the domain of real-time communication, Packet Loss Concealment (PLC) strategies are crucial for mitigating the adverse effects of packet loss and delays, thereby preserving the integrity of speech transmission. Existing methods, while effective to all degrees, often fall short in scenarios characterized by prolonged packet loss, where the quality of speech recovery deteriorates with increased loss duration. Our study introducesSemantic-Aware Speech Encoding, a novel encoding approach that integrates semantic information with audio to improve PLC. By leveraging advances in neural audio coding, our method efficiently extracts and integrates semantic cues alongside audio signals, facilitating their joint transmission over the network. This semantically enriched methodology concurrently facilitating enhanced accuracy in the receiver’s reconstruction of lost speech elements. Experiment results demonstrate its effectiveness, especially in long burst packet loss scenarios, highlighting its reliability and significant impact on PLC advancements.