Audio & Speech Processing

Yuxiang Wang, Hongyu Liu, Yijiang Xu et al.

For researchers and developers of speech language models, this benchmark exposes a pervasive speech grounding gap where models recognize social norms in text but fail to apply them when cues are grounded in speech.

Sreyan Ghosh, Arushi Goel, Kaousheik Jayakumar et al.

This work advances open-source audio-language models for researchers and practitioners needing robust understanding of speech, sound, and music, with strong real-world generalization.

Guanrou Yang, Tian Tan, Qian Chen et al.

For speech AI researchers, WavCube provides a unified representation that bridges the gap between semantic and acoustic features, enabling a single model for both understanding and generation tasks.

Ke-Han Lu, Szu-Wei Fu, Chao-Han Huck Yang et al.

This work provides empirical grounding for understanding LLMs in audio research, addressing a gap in knowledge for researchers and practitioners in audio AI.

Wenxi Chen, Dongya Jia, Yushen Chen et al.

This work addresses the information loss and non-end-to-end training issues in latent-based TTS by directly modeling raw waveforms, offering a new direction for end-to-end speech generation.

Han Zhu, Lingxuan Ye, Wei Kang et al.

This work addresses the challenge of creating a massive multilingual TTS system for broad language coverage, representing a significant advancement rather than an incremental improvement.

Zhen Ye, Xu Tan, Aoxiong Yin et al.

This work improves the quality and efficiency of talking head synthesis by addressing the suboptimal entanglement of high-level semantics and low-level details in existing joint generation models.

Jingyu Lu, Yuhan Wang, Fan Zhuo et al.

This work addresses evaluation challenges for spoken dialogue systems, offering a novel benchmark and model to better assess conversational quality, though it is incremental in advancing existing reward modeling approaches.

Zhifei Xie, Zihang Liu, Ze An et al.

This work addresses the need for a single model that can handle multiple streaming audio tasks (e.g., voice chatting, ASR) in real time, unifying capabilities that were previously separate.

Kai-Wei Chang, Wei-Chih Chen, En-Pei Hu et al.

This addresses a practical limitation for real-world spoken language systems like voice assistants, where controlling response duration can enhance interaction quality, though it is an incremental improvement.

Yi Su, Jisheng Bai, Qisheng Xu et al.

This is an incremental survey that helps researchers and practitioners in audio-centric AI by summarizing existing technologies and providing references for practical applications.

Yadong Niu, Tianzi Wang, Heinrich Dinkel et al.

This addresses the need for better training data to develop versatile large audio-language models for general audio understanding, representing an incremental improvement through a new dataset.

Tingle Li, Siddharth Gururani, Kevin J. Shih et al.

For researchers in video-to-audio generation, this work highlights the need to move beyond perceptual quality toward learning physical processes from pixels, though it is incremental in proposing a new evaluation benchmark.

Yuejiao Wang, Zihao Ji, Pengfei Cai et al.

This work addresses the lack of temporally varying control in neural song generation, providing a method for users to specify local musical attributes aligned with song segments.

Jing Peng, Junhao Du, Chenghao Wang et al.

This framework significantly improves comparability and reproducibility for researchers and developers working on deployment-oriented speech understanding systems.

Kangxiang Xia, Bingshen Mu, Xian Shi et al.

This work addresses the problem of natural full-duplex interaction for spoken dialogue systems, offering a significant improvement over existing methods.

Heinrich Dinkel, Gang Li, Jizhong Liu et al.

This work addresses the need for transparent and reproducible audio-language models for researchers and practitioners, though it is incremental as it builds on existing open-source components and datasets.

Shikhar Bharadwaj, Chin-Jou Li, Kwanghee Choi et al.

Provides an empirical recipe for universal phone recognition, benefiting multilingual and low-resource speech processing.

Jiale Qian, Hao Meng, Tian Zheng et al.

It addresses the need for practical, flexible SVS systems in real-world production workflows, though it appears incremental by building on existing SVS methods.

Chenshuang Zhang, Kyeong Seon Kim, Chengxin Liu et al.

For researchers developing audio-visual LLMs, this work identifies a fundamental limitation in cross-modality understanding between speech and vision, highlighting the need for speech-grounded video comprehension.