Research map / Image generation / Autoregressive and latent generation
SACodec: Asymmetric Quantization with Semantic Anchoring for Low-Bitrate High-Fidelity Neural Speech Codecs
Zhongren Dong, Bin Wang, Jing Han, Haotian Guo, Xiaojun Mo, Yimin Cao, Zixing Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add semantic anchor Q1 (fixed mHuBERT codebook) (ablation) | SACodec · ARCH compressed domain | mean accuracy: .3494 -> .4809 | Improved |
| use fixed mHuBERT codebook instead of learnable VQ (ablation) | SACodec · ARCH compressed domain | mean accuracy: .3356 -> .4809 | Improved |
| use fixed mHuBERT codebook instead of learnable VQ (ablation) | SACodec · LibriTTS test-clean | UTMOS: 3.9051 -> 4.0373 | Improved |
| use larger semantic anchor K1=1000 instead of 500 (ablation) | SACodec · LibriTTS test-clean | UTMOS: 3.8610 -> 4.0373 | Improved |
| add residual quantizer Q2 (SimVQ) (ablation) | SACodec · LibriTTS test-clean | PESQ: 2.3504 -> 2.6937 | Improved |
| add residual quantizer Q2 (SimVQ) (ablation) | SACodec · ARCH compressed domain | mean accuracy: .5065 -> .4809 | Hurt |
| double residual codebook to K2=2048 (ablation) | SACodec · LibriTTS test-clean | UTMOS: 4.0373 -> 4.0402 | No clear effect |
| codebook-space projection instead of direct lookup to fixed semantic codebook (ablation) | SACodec · LibriTTS test-clean | UTMOS: 3.65 -> 4.04 | Improved |
Closest papers by meaning
- ALMTokenizer: A Low-bitrate and Semantic-rich Audio Codec Tokenizer for Audio Language Modeling (ICML 2025 · similarity 0.89)
- WavTokenizer: an Efficient Acoustic Discrete Codec Tokenizer for Audio Language Modeling (ICLR 2025 · similarity 0.89)
- Scaling Transformers for Low-Bitrate High-Quality Speech Coding (ICLR 2025 · similarity 0.88)
- FocalCodec: Low-Bitrate Speech Coding via Focal Modulation Networks (NeurIPS 2025 · similarity 0.88)
- Codec Does Matter: Exploring the Semantic Shortcoming of Codec for Audio Language Model (AAAI 2025 · similarity 0.87)
- CLaM-TTS: Improving Neural Codec Language Model for Zero-Shot Text-to-Speech (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
