arXiv:2611.14043v1 Announce Type: new Abstract: Despite recent advances in neural speech codecs, achieving highfidelity speech reconstruction at low bitrates remains a formidable challenge. To address this limitation, we propose ESDCodec, a speech codec that integrates a thoroughly enhanced semantic quantizer and a conditioned decoder network. Specifically, we employ a randomly initialized and frozen codebook, followed by a lightweight projector, to encode semantic details entirely within a linear space while enhancing codebook utilization. To further improve perceptual quality, we design a condition network that explicitly injects prior subband knowledge into the upsampling decoder. Taking the de-quantized feature as input, this network predicts subband signals, thereby providing fine-grained guidance for waveform reconstruction. Extensive experimental results show that ESDCodec achieves superior reconstruction performance at an ultra-low bitrate of 0.85kbps. For LLM-based speech generation tasks, ESDCodec also consistently outperforms existing codec models.