论文解读
并行相加代替串行求余：PACodec 如何用小码本做低码率语音编码
PACodec 用并行加性量化，在 LibriTTS 1.4 kbps 与 VCTK 4.2 kbps 上取得接近较高码率基线的客观表现；主观 ABX 仅在 LibriTTS 上进行，对多数 2 kbps 基线未检出显著偏好差异，并非等效证明，分支解耦也仍只是消融显示的潜力。
生成对抗网络
论文解读
不看未来也能补高频：StreamWSR 的因果波形超分
StreamWSR 把低采样语音先上采样再用全因果波形网络预测残差补高频，在 VCTK 三种带宽扩展设置下以 9M 参数和 2G FLOPs 达到与非流式基线相当的失真与可懂度，并用消融证明了帧级压缩与频谱判别器的作用。
论文解读
在一条潜流里同时生成两条音轨：解耦语义与声学后的少步分离
音乐源分离 | 5.3/10
论文解读
既要指向性，又要去混响：一次前向如何重建干净的虚拟方向麦克风
空间音频 | 5.3/10
论文解读
让声码器也理解方向：CSAVocoder 把空间线索留在最后一公里
空间音频 | 7.6/10
论文解读
Architecture and Affordances of PLAUD: Performative Latents and Unsupervised DDSP
音乐生成 | 5.0/10
论文解读
LILAC: An Idempotent Neural Speech Codec
语音编码 | 8.1/10
论文解读
Efficient Audio Enhancement with a Differentiable Psychoacoustic Loss
音频超分辨 | 8.2/10
论文解读
Revisiting Vocos: That Phasiness Business in Time-Frequency Neural Vocoding
语音合成 | 7.1/10
论文解读
MusicDET: Zero-Shot AI-Generated Music Detection
音频伪造检测 | 6.1/10
论文解读
Query-Based Asymmetric Modeling with Decoupled Input–Output Rates for Speech Restoration
语音增强 | 7.1/10
论文解读
OLIVE: View-Augmented Latent Prediction with Waveform Reconstruction for Speech SSL
语音识别 | 7.5/10
论文解读
Joint Residual Reweighting for Classifier Free Guidance in Flow-Matching Zero-Shot TTS
语音合成 | 7.5/10
论文解读
DTT-BSR+: A Generative-Regression Cascade for Music Source Restoration
生成对抗网络 | 8/10
论文解读
Exploiting Neural Audio Codec Latents for Adversarial Audio Attacks
生成对抗网络 | 7.2/10
论文解读
PhysDrift: Bridging the Embodiment Gap in Humanoid Co-Speech Motion Generation
语音合成 | 7.4/10
论文解读
Zero-VC: Zero-Lookahead Streaming Voice Conversion via Speaker Anonymization
语音转换 | 6.1/10
论文解读
QC-GAN: A Parameter-Efficient Quaternion Conformer GAN for High-Fidelity Speech Enhancement
语音增强 | 7.1/10
论文解读
Unified Audio Generation and Editing via Joint Condition Modeling and Progressive Training
音频生成 | 8.7/10
论文解读
Feature-Aligned Speech Watermarking for Robustness to Reconstruction Distortions
鲁棒性 | 7.1/10