[ { "arxiv": "2503.01710", "version": "2503.01710v1", "title": "Spark-TTS: An Efficient LLM-Based Text-to-Speech Model with Single-Stream Decoupled Speech Tokens", "authors": [ "Xinsheng Wang", "Mingqi Jiang", "Ziyang Ma", "Ziyu Zhang", "Songxiang Liu", "Linqin Li", "Zheng Liang", "Qixi Zheng", "Rui Wang", "Xiaoqin Feng", "Weizhen Bian", "Zhen Ye", "Sitong Cheng", "Ruibin Yuan", "Zhixian Zhao", "Xinfa Zhu", "Jiahao Pan", "Liumeng Xue", "Pengcheng Zhu", "Yunlin Chen", "Zhifei Li", "Xie Chen", "Lei Xie", "Yike Guo", "Wei Xue" ], "published": "2025-03-03T16:23:10Z", "updated": "2025-03-03T16:23:10Z", "field": "synthesis", "url": "https://arxiv.org/abs/2503.01710v1", "group": "S1", "label": "Spark-TTS", "problem": "Separate speaker identity from changing speech content.", "approach": "A global speaker representation conditions reconstruction; a single token stream carries changing speech information.", "boundary": "A useful 2025 anchor; separation alone does not establish a new contribution.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S01" }, { "arxiv": "2506.21619", "version": "2506.21619v2", "title": "IndexTTS2: A Breakthrough in Emotionally Expressive and Duration-Controlled Auto-Regressive Zero-Shot Text-to-Speech", "authors": [ "Siyi Zhou", "Yiquan Zhou", "Yi He", "Xun Zhou", "Jinchao Wang", "Wei Deng", "Jingchen Shu" ], "published": "2025-06-23T08:33:40Z", "updated": "2025-09-03T10:46:35Z", "field": "synthesis", "url": "https://arxiv.org/abs/2506.21619v2", "group": "S1", "label": "IndexTTS2", "problem": "Control emotion and duration without losing the requested voice.", "approach": "Separate emotion and identity prompts; add token-count duration control and text-based emotional instructions.", "boundary": "Judge duration, intelligibility, emotion and speaker identity separately.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S02" }, { "arxiv": "2512.13251", "version": "2512.13251v3", "title": "DisCo-Speech: Controllable Zero-Shot Speech Generation with A Disentangled Speech Codec", "authors": [ "Tao Li", "Wenshuo Ge", "Zhichao Wang", "Zihao Cui", "Yong Ma", "Yingying Gao", "Chao Deng", "Shilei Zhang", "Junlan Feng" ], "published": "2025-12-15T12:06:17Z", "updated": "2026-01-04T02:19:32Z", "field": "synthesis", "url": "https://arxiv.org/abs/2512.13251v3", "group": "S1", "label": "DisCo-Speech", "problem": "Transfer style without copying the reference speaker.", "approach": "A disentangled codec separates content, prosody and timbre for controlled zero-shot generation.", "boundary": "Independent control needs tests on previously unseen combinations.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S03" }, { "arxiv": "2603.08823", "version": "2603.08823v2", "title": "Fish Audio S2 Technical Report", "authors": [ "Shijia Liao", "Yuxuan Wang", "Songting Liu", "Yifan Cheng", "Ruoyi Zhang", "Tianyu Li", "Shidong Li", "Yisheng Zheng", "Xingwei Liu", "Qingzheng Wang", "Zhizhuo Zhou", "Jiahua Liu", "Xin Chen", "Dawei Han" ], "published": "2026-03-09T18:34:33Z", "updated": "2026-03-11T08:59:27Z", "field": "synthesis", "url": "https://arxiv.org/abs/2603.08823v2", "group": "S1", "label": "Fish Audio S2", "problem": "Follow expressive instructions in speech generation.", "approach": "A dual autoregressive architecture supports expressive, multi-speaker and multi-turn synthesis.", "boundary": "Treat report comparisons as protocol-specific evidence.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S04" }, { "arxiv": "2605.24618", "version": "2605.24618v1", "title": "FC-TTS: Style and Timbre Control in Zero-Shot Text-to-Speech with Disentangled Speech Representations", "authors": [ "Yoonhyung Lee", "Hyunsin Park", "Jinhwan Park", "Jinkyu Lee" ], "published": "2026-05-23T15:01:28Z", "updated": "2026-05-23T15:01:28Z", "field": "synthesis", "url": "https://arxiv.org/abs/2605.24618v1", "group": "S1", "label": "FC-TTS", "problem": "Choose style and voice from different recordings.", "approach": "Separate reference representations and consistency objectives support independent style and timbre control.", "boundary": "A changed style can affect perceived speaker identity; evaluate both together.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S05" }, { "arxiv": "2608.24168", "version": "2608.24168v2", "title": "FireRedAudio: A General-Purpose Audio Language Model with Decoupled Continuous Representations for Understanding and Generation", "authors": [ "Feiyu Shen", "Fenglong Xie", "Junjie Li", "Kun Xie", "Lei Xie", "Xu Tang", "Xuelong Geng", "Yan Jia", "Yao Hu", "Yichen Han", "Yichen Wu", "Ziqi Dai", "Junjie Chen", "Kai Huang", "Manzhen Wei", "Yixuan Li" ], "published": "2026-08-25T07:37:15Z", "updated": "2026-08-27T03:36:15Z", "field": "synthesis", "url": "https://arxiv.org/abs/2608.24168v2", "group": "S1", "label": "FireRedAudio", "problem": "Support synthesis and speech editing in a unified audio model.", "approach": "Use separate continuous input paths for understanding and reconstructible generation; condition a flow model for speech.", "boundary": "Included for synthesis and editing; its broader understanding tasks are outside this map.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S06" }, { "arxiv": "2610.18010", "version": "2610.18010v2", "title": "Fish Audio Drama 3 Technical Report", "authors": [ "Dawei Han", "Shidong Li", "Shijia Liao", "Jiahua Liu", "Songting Liu", "Qingzheng Wang", "Yuxuan Wang", "Xin Chen", "Yifan Cheng", "Tianyu Li", "Gangin Park", "Xuehao Yang", "Chang Zeng", "Haoyang Zhang", "Ruoyi Zhang", "Xiangyu Zhang" ], "published": "2026-10-13T09:06:28Z", "updated": "2026-10-14T09:28:32Z", "field": "synthesis", "url": "https://arxiv.org/abs/2610.18010v2", "group": "S1", "label": "Fish Audio Drama 3", "problem": "Generate and revise expressive multi-speaker scenes.", "approach": "Scale speech modeling, predict multiple codebooks and use human preference training.", "boundary": "The evaluation includes regeneration of failed outputs; single-attempt reliability needs separate testing.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S07" }, { "arxiv": "2602.00594", "version": "2602.00594v2", "title": "Kanade: A Simple Disentangled Tokenizer for Spoken Language Modeling", "authors": [ "Zhijie Huang", "Stephen McIntosh", "Daisuke Saito", "Nobuaki Minematsu" ], "published": "2026-01-31T08:20:39Z", "updated": "2026-05-05T13:18:46Z", "field": "synthesis", "url": "https://arxiv.org/abs/2602.00594v2", "group": "S2", "label": "Kanade", "problem": "Keep pronunciation and rhythm while separating stable acoustic information.", "approach": "A single discrete stream carries changing speech features; a global continuous vector carries acoustic constants.", "boundary": "Full-text caveat: the presented tokenizer is not natively streaming.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S08" }, { "arxiv": "2602.06602", "version": "2602.06602v1", "title": "Scaling Speech Tokenizers with Diffusion Autoencoders", "authors": [ "Yuancheng Wang", "Zhenyu Tang", "Yun Wang", "Arthur Hinsvark", "Yingru Liu", "Yinghao Li", "Kainan Peng", "Junyi Ao", "Mingbo Ma", "Mike Seltzer", "Qing He", "Xubo Liu" ], "published": "2026-02-06T10:57:41Z", "updated": "2026-02-06T10:57:41Z", "field": "synthesis", "url": "https://arxiv.org/abs/2602.06602v1", "group": "S2", "label": "SiTok", "problem": "Compress speech without sacrificing useful generation information.", "approach": "Scale a diffusion-autoencoder tokenizer with supervised semantic information and low-rate speech tokens.", "boundary": "Reconstruction quality alone does not measure downstream generation quality.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S09" }, { "arxiv": "", "version": "", "title": "XY-Tokenizer: Mitigating the Semantic-Acoustic Conflict in Low-Bitrate Speech Codecs", "authors": [ "Yitian Gong", "Luozhijie Jin", "Kuangwei Chen", "Dong Zhang", "Ruifan Deng", "Xiaogui Yang", "Xin Zhang", "Zhaoye Fei", "Qinyuan Cheng", "Shimin Li", "Xipeng Qiu" ], "published": "2026-07", "updated": "", "url": "https://aclanthology.org/2026.acl-long.423/", "field": "synthesis", "group": "S2", "label": "XY-Tokenizer", "problem": "Balance linguistic information and acoustic detail at low bitrate.", "approach": "Use staged multitask training to align discrete speech features with text while preserving reconstruction.", "boundary": "Evaluate representation quality using downstream understanding and generation, not bitrate alone.", "evidence": "Official ACL 2026 record and abstract checked.", "venue": "ACL 2026", "doi": "10.18653/v1/2026.acl-long.423", "ref": "S10" }, { "arxiv": "2608.08286", "version": "2608.08286v1", "title": "ReLMCodec: Designing Predictable Speech Tokens from Pre-Quantization Phoneme Structure", "authors": [ "Zixiang Wan", "Xusheng Yang", "Zheng Wang", "Peiji Yang" ], "published": "2026-08-08T18:35:10Z", "updated": "2026-08-08T18:35:10Z", "field": "synthesis", "url": "https://arxiv.org/abs/2608.08286v1", "group": "S2", "label": "ReLMCodec", "problem": "Make speech tokens easier for a language model to predict.", "approach": "Organize pre-quantization features using phoneme structure to improve token predictability.", "boundary": "Compare language-model difficulty alongside speech reconstruction quality.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S11" }, { "arxiv": "2609.30975", "version": "2609.30975v1", "title": "A Comprehensive Study of Content Representations for Speech Synthesis", "authors": [ "Diego Torres", "Axel Roebel", "Nicolas Obin" ], "published": "2026-09-25T08:20:56Z", "updated": "2026-09-25T08:20:56Z", "field": "synthesis", "url": "https://arxiv.org/abs/2609.30975v1", "group": "S2", "label": "Content representation study", "problem": "Measure what content representations actually retain.", "approach": "Compare representations under one generative framework along content, speaker and prosody dimensions.", "boundary": "The study links disentanglement to both objective and information capacity.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S12" }, { "arxiv": "2610.01492", "version": "2610.01492v1", "title": "Q-SPT: Learnable Query-Based Compression for Low-Frame-Rate Speech Tokenization", "authors": [ "Jeeyoung Yun", "Seohwan Yun", "Sungwoong Kim" ], "published": "2026-10-01T11:35:01Z", "updated": "2026-10-01T11:35:01Z", "field": "synthesis", "url": "https://arxiv.org/abs/2610.01492v1", "group": "S2", "label": "Q-SPT", "problem": "Reduce speech token rate while keeping useful detail.", "approach": "Learn query-based compression for semantic and acoustic features.", "boundary": "A lower frame rate needs controlled quality and downstream-task comparisons.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S13" }, { "arxiv": "2610.29150", "version": "2610.29150v1", "title": "Where Does Timbre Hide? Localizing Speaker Information in Speech Signals", "authors": [ "Yiquan Zhou", "Yixuan Wang", "Xiaozhe Qi", "Xin Gao", "Shihao Li" ], "published": "2026-10-21T03:35:42Z", "updated": "2026-10-21T03:35:42Z", "field": "synthesis", "url": "https://arxiv.org/abs/2610.29150v1", "group": "S2", "label": "Where Does Timbre Hide?", "problem": "Locate residual speaker information in signals and tokens.", "approach": "Probe frequency bands, signal components, model layers and tokenizers using speaker identity and listener tests.", "boundary": "Speaker identity is a measurable proxy for timbre, not its complete definition.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S14" }, { "arxiv": "2505.17589", "version": "2505.17589v2", "title": "CosyVoice 3: Towards In-the-wild Speech Generation via Scaling-up and Post-training", "authors": [ "Zhihao Du", "Changfeng Gao", "Yuxuan Wang", "Fan Yu", "Tianyu Zhao", "Hao Wang", "Xiang Lv", "Hui Wang", "Chongjia Ni", "Xian Shi", "Keyu An", "Guanrou Yang", "Yabin Li", "Yanni Chen", "Zhifu Gao", "Qian Chen", "Yue Gu", "Mengzhe Chen", "Yafeng Chen", "Shiliang Zhang", "Wen Wang", "Jieping Ye" ], "published": "2025-05-23T07:55:21Z", "updated": "2025-05-27T07:48:34Z", "field": "synthesis", "url": "https://arxiv.org/abs/2505.17589v2", "group": "S3", "label": "CosyVoice 3", "problem": "Improve synthesis on difficult real-world text and references.", "approach": "Combine scale, a multitask speech tokenizer and post-training feedback.", "boundary": "Useful system-level evidence; isolate the contribution of data, objectives and model scale.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S15" }, { "arxiv": "2601.15621", "version": "2601.15621v1", "title": "Qwen3-TTS Technical Report", "authors": [ "Hangrui Hu", "Xinfa Zhu", "Ting He", "Dake Guo", "Bin Zhang", "Xiong Wang", "Zhifang Guo", "Ziyue Jiang", "Hongkun Hao", "Zishan Guo", "Xinyu Zhang", "Pei Zhang", "Baosong Yang", "Jin Xu", "Jingren Zhou", "Junyang Lin" ], "published": "2026-01-22T03:51:43Z", "updated": "2026-01-22T03:51:43Z", "field": "synthesis", "url": "https://arxiv.org/abs/2601.15621v1", "group": "S3", "label": "Qwen3-TTS", "problem": "Support controllable multilingual synthesis with efficient speech generation.", "approach": "Present alternative tokenization and generation tracks, including causal speech decoding.", "boundary": "Token rate, codebook count and end-to-end latency are different quantities.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S16" }, { "arxiv": "2606.06928", "version": "2606.06928v1", "title": "VoxCPM2 Technical Report", "authors": [ "Yixuan Zhou", "Guoyang Zeng", "Xin Liu", "Xiang Li", "Renjie Yu", "Jiancheng Gui", "Jiaheng Wu", "Ziyang Wang", "Xudong Shen", "Runchuan Ye", "Zhisheng Zhang", "Jiuyang Zhou", "Bingsong Bai", "Weiyue Sun", "Mengyuan Deng", "Qundong Shi", "Zhiyong Wu", "Zhiyuan Liu" ], "published": "2026-06-05T05:43:15Z", "updated": "2026-06-05T05:43:15Z", "field": "synthesis", "url": "https://arxiv.org/abs/2606.06928v1", "group": "S3", "label": "VoxCPM2", "problem": "Model speech directly in continuous acoustic representations.", "approach": "Predict acoustic patches autoregressively and generate local detail with diffusion.", "boundary": "It avoids an external discrete token stream; internal quantization still exists.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S17" }, { "arxiv": "2606.09234", "version": "2606.09234v1", "title": "End-to-End Training for Discrete Token LLM based TTS System", "authors": [ "Changfeng Gao", "Yong Ren", "Jun Yuan", "Ye Bai", "Zhao You", "ShiDong Shang" ], "published": "2026-06-08T09:07:23Z", "updated": "2026-06-08T09:07:23Z", "field": "synthesis", "url": "https://arxiv.org/abs/2606.09234v1", "group": "S3", "label": "End-to-end discrete-token TTS", "problem": "Reduce mismatch between separately trained synthesis components.", "approach": "Train tokenizer representations with reconstruction, prediction and recognition objectives; then use downstream feedback for the language model.", "boundary": "This is staged joint optimization; not every parameter is updated in every stage.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S18" }, { "arxiv": "2607.23938", "version": "2607.23938v1", "title": "Qwen-Audio-3.0-TTS: Freely Controllable and Highly Robust Speech Synthesis with Multi-Stage Training Paradigm", "authors": [ "Bajian Xiang", "Cheng Wen", "Han Zhao", "Hao Wang", "Haoxu Wang", "Jiawei Jin", "Jiayan Cui", "Jie Chen", "Mengxi Nie", "Tianyu Zhao", "Weiqin Li", "Xiang Lv", "Xiangang Li", "Yang Xiang", "Yang Zhou" ], "published": "2026-07-27T02:30:15Z", "updated": "2026-07-27T02:30:15Z", "field": "synthesis", "url": "https://arxiv.org/abs/2607.23938v1", "group": "S3", "label": "Qwen-Audio-3.0-TTS", "problem": "Improve control and robustness through coordinated training.", "approach": "Use a multi-stage recipe that combines language-model training, flow-model training and reinforcement learning.", "boundary": "Separate improvements in pronunciation, control, reference robustness and long-text behavior.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S19" }, { "arxiv": "2608.11593", "version": "2608.11593v1", "title": "Luna-TTS Family Technical Report", "authors": [ "Feng Yin", "Shuai Shi", "Junjie Zheng", "Kechenying Zhou", "Yiqiu Wang", "Chenyang He", "Qiuhua Jiang", "Mengxiao Bi", "Yanmin Qian", "Mingxin Chen", "Xun Gong", "Tianteng Gu", "Bing Han", "Peng Jiang", "Chenda Li", "Haiyang Sun", "Han Wang", "Wei Wang", "Yi Wang", "Leying Zhang", "Wangyou Zhang", "Chushu Zhou" ], "published": "2026-08-12T03:07:07Z", "updated": "2026-08-12T03:07:07Z", "field": "synthesis", "url": "https://arxiv.org/abs/2608.11593v1", "group": "S3", "label": "Luna-TTS", "problem": "Reduce sequential decoding while retaining controllable generation.", "approach": "Denoise a discrete token grid in parallel; use blockwise generation for the realtime variant.", "boundary": "The realtime variant uses 1.28-second blocks; reported warm latency is protocol-specific.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S20" }, { "arxiv": "2609.16514", "version": "2609.16514v1", "title": "The Evolving Bottleneck in Speech Generation: Interface Co-design and Staged Alignment from CosyVoice to Qwen-Audio-3.0-TTS", "authors": [ "Qian Chen", "Xiangang Li", "Xiang Lv", "Han Zhao", "Tianyu Zhao" ], "published": "2026-09-15T02:04:54Z", "updated": "2026-09-15T02:04:54Z", "field": "synthesis", "url": "https://arxiv.org/abs/2609.16514v1", "group": "S3", "label": "The Evolving Bottleneck", "problem": "Explain why improving one synthesis component can stop helping.", "approach": "A technical analysis traces interface design and staged alignment from CosyVoice to Qwen-Audio-3.0-TTS.", "boundary": "Background synthesis of one model lineage; not an independent new benchmark or universal proof.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S21" }, { "arxiv": "2610.28035", "version": "2610.28035v1", "title": "LoCA: Local Credit Assignment for Reinforcement Learning of Autoregressive Flow-Matching TTS", "authors": [ "Jie Liu", "Junyu Zhou", "Junhao Zhong", "WeiSen Tang", "Rukai Cai", "Haoyue Zhan", "Ruiqi Xing", "Ling Long" ], "published": "2026-10-20T11:23:47Z", "updated": "2026-10-20T11:23:47Z", "field": "synthesis", "url": "https://arxiv.org/abs/2610.28035v1", "group": "S3", "label": "LoCA", "problem": "Give local feedback when only part of an utterance is wrong.", "approach": "Use ASR alignment to redistribute rewards across acoustic frames and constrain positive and negative regression updates.", "boundary": "Experiments use VoxCPM2; transfer to other synthesis architectures remains a research question.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S22" }, { "arxiv": "2502.18924", "version": "2502.18924v4", "title": "MegaTTS 3: Sparse Alignment Enhanced Latent Diffusion Transformer for Zero-Shot Speech Synthesis", "authors": [ "Ziyue Jiang", "Yi Ren", "Ruiqi Li", "Shengpeng Ji", "Boyang Zhang", "Zhenhui Ye", "Chen Zhang", "Bai Jionghao", "Xiaoda Yang", "Jialong Zuo", "Yu Zhang", "Rui Liu", "Xiang Yin", "Zhou Zhao" ], "published": "2025-02-26T08:22:00Z", "updated": "2025-03-28T05:34:33Z", "field": "synthesis", "url": "https://arxiv.org/abs/2502.18924v4", "group": "S4", "label": "MegaTTS 3", "problem": "Balance reliable text alignment with natural speech.", "approach": "Sparse alignment guides latent diffusion; additional conditioning controls accent intensity.", "boundary": "Use difficult sentences to test alignment, rather than relying only on average naturalness.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S23" }, { "arxiv": "2506.13053", "version": "2506.13053v3", "title": "ZipVoice: Fast and High-Quality Zero-Shot Text-to-Speech with Flow Matching", "authors": [ "Han Zhu", "Wei Kang", "Zengwei Yao", "Liyong Guo", "Fangjun Kuang", "Zhaoqing Li", "Weiji Zhuang", "Long Lin", "Daniel Povey" ], "published": "2025-06-16T02:48:17Z", "updated": "2025-08-07T03:12:26Z", "field": "synthesis", "url": "https://arxiv.org/abs/2506.13053v3", "group": "S4", "label": "ZipVoice", "problem": "Make high-quality zero-shot synthesis cheaper to run.", "approach": "Use compact Zipformer networks, initial text alignment and flow distillation to reduce inference cost.", "boundary": "Speed comparisons depend on model, hardware, sampling settings and quality targets.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S24" }, { "arxiv": "2507.09318", "version": "2507.09318v2", "title": "ZipVoice-Dialog: Non-Autoregressive Spoken Dialogue Generation with Flow Matching", "authors": [ "Han Zhu", "Wei Kang", "Liyong Guo", "Zengwei Yao", "Fangjun Kuang", "Weiji Zhuang", "Zhaoqing Li", "Zhifeng Han", "Dong Zhang", "Xin Zhang", "Xingchen Song", "Lingxuan Ye", "Long Lin", "Daniel Povey" ], "published": "2025-07-12T15:18:47Z", "updated": "2026-04-14T14:21:41Z", "field": "synthesis", "url": "https://arxiv.org/abs/2507.09318v2", "group": "S4", "label": "ZipVoice-Dialog", "problem": "Synthesize scripted dialogue with correct speaker turns.", "approach": "Use alignment curriculum and speaker-turn embeddings in non-autoregressive flow matching; release OpenDialog.", "boundary": "Scripted multi-speaker synthesis does not establish interactive full-duplex dialogue.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S25" }, { "arxiv": "2508.19205", "version": "2508.19205v1", "title": "VibeVoice Technical Report", "authors": [ "Zhiliang Peng", "Jianwei Yu", "Wenhui Wang", "Yaoyao Chang", "Yutao Sun", "Li Dong", "Yi Zhu", "Weijiang Xu", "Hangbo Bao", "Zehua Wang", "Shaohan Huang", "Yan Xia", "Furu Wei" ], "published": "2025-08-26T17:09:12Z", "updated": "2025-08-26T17:09:12Z", "field": "synthesis", "url": "https://arxiv.org/abs/2508.19205v1", "group": "S4", "label": "VibeVoice", "problem": "Maintain voices and context across long multi-speaker recordings.", "approach": "Combine low-rate continuous representations with autoregressive latent generation and a diffusion head.", "boundary": "Long generation capability does not itself prove consistent quality for the whole recording.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S26" }, { "arxiv": "2509.15969", "version": "2509.15969v2", "title": "VoXtream: Full-Stream Text-to-Speech with Extremely Low Latency", "authors": [ "Nikita Torgashov", "Gustav Eje Henter", "Gabriel Skantze" ], "published": "2025-09-19T13:26:46Z", "updated": "2026-01-26T16:25:29Z", "field": "synthesis", "url": "https://arxiv.org/abs/2509.15969v2", "group": "S4", "label": "VoXtream", "problem": "Start speaking before the full input text is available.", "approach": "Use incremental text processing and bounded future context in a full-stream TTS pipeline.", "boundary": "Measure quality against available text lookahead and first-audio latency.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S27" }, { "arxiv": "2603.13518", "version": "2603.13518v1", "title": "VoXtream2: Full-stream TTS with dynamic speaking rate control", "authors": [ "Nikita Torgashov", "Gustav Eje Henter", "Gabriel Skantze" ], "published": "2026-03-13T18:49:25Z", "updated": "2026-03-13T18:49:25Z", "field": "synthesis", "url": "https://arxiv.org/abs/2603.13518v1", "group": "S4", "label": "VoXtream2", "problem": "Control speaking rate during full-stream synthesis.", "approach": "Extend full-stream generation with dynamic rate control and prompt masking during training.", "boundary": "Test rapidly changing text and rate requests as well as fixed sentences.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S28" }, { "arxiv": "2603.18090", "version": "2603.18090v2", "title": "MOSS-TTS Technical Report", "authors": [ "Yitian Gong", "Botian Jiang", "Yiwei Zhao", "Yucheng Yuan", "Kuangwei Chen", "Yaozhou Jiang", "Cheng Chang", "Dong Hong", "Mingshu Chen", "Ruixiao Li", "Yiyang Zhang", "Yang Gao", "Hanfu Chen", "Ke Chen", "Songlin Wang", "Xiaogui Yang", "Yuqian Zhang", "Kexin Huang", "ZhengYuan Lin", "Kang Yu", "Ziqi Chen", "Jin Wang", "Zhaoye Fei", "Qinyuan Cheng", "Shimin Li", "Xipeng Qiu" ], "published": "2026-03-18T09:08:06Z", "updated": "2026-03-20T08:16:20Z", "field": "synthesis", "url": "https://arxiv.org/abs/2603.18090v2", "group": "S4", "label": "MOSS-TTS", "problem": "Scale discrete-token speech generation for complex long contexts.", "approach": "A speech language-model system targets multilingual, controllable and long-form generation.", "boundary": "Assess content errors, voice consistency and long-context stability separately.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S29" }, { "arxiv": "2605.28618", "version": "2605.28618v1", "title": "Comprehensive Benchmarking of Long-Form Speech Generation in Diverse Scenarios", "authors": [ "Changhao Pan", "Rui Yang", "Han Wang", "Zhuan Zhou", "Xuming He", "Wenxiang Guo", "Ziyue Jiang", "Ruiqi Li", "Yu Zhang", "Chenyuhao Wen", "Ke Lei", "Xiang Yin", "Jingyu Lu", "Zhiyuan Zhu", "Zhou Zhao" ], "published": "2026-05-27T15:28:15Z", "updated": "2026-05-27T15:28:15Z", "field": "synthesis", "url": "https://arxiv.org/abs/2605.28618v1", "group": "S4", "label": "SwanBench-Speech", "problem": "Evaluate long-form quality beyond short-utterance scores.", "approach": "Benchmark diverse scenarios across acoustics, semantics and expressiveness using seven metrics.", "boundary": "Automated scores require human validation for discourse-level expression and coherence.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S30" }, { "arxiv": "2609.08703", "version": "2609.08703v2", "title": "TontaubeV1: Streaming Text-to-Speech with Hierarchical Codec Modeling and Bounded Context", "authors": [ "Fritz Cremer", "Jonathan Cremer" ], "published": "2026-09-08T13:03:46Z", "updated": "2026-10-13T15:24:21Z", "field": "synthesis", "url": "https://arxiv.org/abs/2609.08703v2", "group": "S4", "label": "TontaubeV1", "problem": "Stream long speech without unbounded context growth.", "approach": "Use hierarchical codec modeling and bounded context for streaming synthesis.", "boundary": "Streaming audio output alone does not prove support for incremental text input.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S31" }, { "arxiv": "2609.09719", "version": "2609.09719v1", "title": "StreamAlign: Streaming Text-Aligned Speech Tokenization", "authors": [ "Kang-wook Kim", "Jinyoung Park", "Jinsoo Kim", "Sehun Lee", "Sang Hoon Woo", "Gunhee Kim" ], "published": "2026-09-09T05:03:12Z", "updated": "2026-09-09T05:03:12Z", "field": "synthesis", "url": "https://arxiv.org/abs/2609.09719v1", "group": "S4", "label": "StreamAlign", "problem": "Build speech tokens without waiting for the complete recording.", "approach": "Combine online alignment, ASR guidance and word-boundary prediction for streaming text-aligned tokens.", "boundary": "A representation study with speech-continuation evidence; not a complete TTS deployment claim.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "S32" }, { "arxiv": "2502.08857", "version": "2502.08857v4", "title": "ASVspoof 5: Design, Collection and Validation of Resources for Spoofing, Deepfake, and Adversarial Attack Detection Using Crowdsourced Speech", "authors": [ "Xin Wang", "Héctor Delgado", "Hemlata Tak", "Jee-weon Jung", "Hye-jin Shim", "Massimiliano Todisco", "Ivan Kukanov", "Xuechen Liu", "Md Sahidullah", "Tomi Kinnunen", "Nicholas Evans", "Kong Aik Lee", "Junichi Yamagishi", "Myeonghun Jeong", "Ge Zhu", "Yongyi Zang", "You Zhang", "Soumi Maiti", "Florian Lux", "Nicolas Müller", "Wangyou Zhang", "Chengzhe Sun", "Shuwei Hou", "Siwei Lyu", "Sébastien Le Maguer", "Cheng Gong", "Hanjie Guo", "Liping Chen", "Vishwanath Singh" ], "published": "2025-02-13T00:15:54Z", "updated": "2025-04-24T22:48:39Z", "field": "detection", "url": "https://arxiv.org/abs/2502.08857v4", "group": "D1", "label": "ASVspoof 5 resources", "problem": "Build a demanding shared test of spoofing and deepfake detection.", "approach": "Release and validate crowdsourced speech resources with diverse synthesis and adversarial conditions.", "boundary": "A resource paper; pair it with the later evaluation report instead of counting two independent benchmarks.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D01" }, { "arxiv": "2509.21597", "version": "2509.21597v2", "title": "AUDDT: A Unified Benchmark Toolkit for Audio and Speech Deepfake Detectors", "authors": [ "Yi Zhu", "Heitor R. Guimarães", "Arthur Pimentel", "Tiago Falk" ], "published": "2025-09-25T21:09:40Z", "updated": "2026-06-03T17:37:13Z", "field": "detection", "url": "https://arxiv.org/abs/2509.21597v2", "group": "D1", "label": "AUDDT", "problem": "Evaluate the same detector across many existing datasets.", "approach": "Provide a unified toolkit and metadata analysis covering 31 datasets in the checked revision.", "boundary": "A broad toolkit does not remove overlap, label or recording-condition biases in its source datasets.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D02" }, { "arxiv": "2510.23096", "version": "2510.23096v1", "title": "TwinShift: Benchmarking Audio Deepfake Detection across Synthesizer and Speaker Shifts", "authors": [ "Jiyoung Hong", "Yoonseo Chung", "Seungyeon Oh", "Juntae Kim", "Jiyoung Lee", "Sookyung Kim", "Hyunsoo Cho" ], "published": "2025-10-27T08:06:07Z", "updated": "2025-10-27T08:06:07Z", "field": "detection", "url": "https://arxiv.org/abs/2510.23096v1", "group": "D1", "label": "TwinShift", "problem": "Separate unseen-speaker effects from unseen-generator effects.", "approach": "Create evaluation splits with disjoint speakers and synthesis systems.", "boundary": "Use the split design to test transfer; avoid treating one aggregate score as universal generalization.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D03" }, { "arxiv": "2601.03944", "version": "2601.03944v3", "title": "ASVspoof 5: Evaluation of Spoofing, Deepfake, and Adversarial Attack Detection Using Crowdsourced Speech", "authors": [ "Xin Wang", "Héctor Delgado", "Nicholas Evans", "Xuechen Liu", "Tomi Kinnunen", "Hemlata Tak", "Kong Aik Lee", "Ivan Kukanov", "Md Sahidullah", "Massimiliano Todisco", "Junichi Yamagishi" ], "published": "2026-01-07T14:01:10Z", "updated": "2026-04-11T14:44:45Z", "field": "detection", "url": "https://arxiv.org/abs/2601.03944v3", "group": "D1", "label": "ASVspoof 5 evaluation", "problem": "Measure detector performance on the shared challenge.", "approach": "Analyze submitted systems, neural-codec attacks, adversarial conditions and evaluation behavior.", "boundary": "Challenge rankings are tied to its data and operating conditions.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D04" }, { "arxiv": "2603.01482", "version": "2603.01482v1", "title": "A SUPERB-Style Benchmark of Self-Supervised Speech Models for Audio Deepfake Detection", "authors": [ "Hashim Ali", "Nithin Sai Adupa", "Surya Subramani", "Hafiz Malik" ], "published": "2026-03-02T05:45:55Z", "updated": "2026-03-02T05:45:55Z", "field": "detection", "url": "https://arxiv.org/abs/2603.01482v1", "group": "D1", "label": "Spoof-SUPERB", "problem": "Compare pretrained speech representations for detection fairly.", "approach": "Evaluate 20 self-supervised models across in-domain, out-of-domain and degraded audio.", "boundary": "Backbone rankings depend on training and evaluation choices.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D05" }, { "arxiv": "2603.02364", "version": "2603.02364v4", "title": "When Spoof Detectors Travel: Evaluation Across 66 Languages in the Low-Resource Language Spoofing Corpus", "authors": [ "Kirill Borodin", "Vasiliy Kudryavtsev", "Maxim Maslov", "Mikhail Gorodnichev", "Grach Mkrtchian" ], "published": "2026-03-02T20:11:06Z", "updated": "2026-06-27T13:26:12Z", "field": "detection", "url": "https://arxiv.org/abs/2603.02364v4", "group": "D1", "label": "LRLspoof", "problem": "Test whether a detector transfers across languages.", "approach": "Evaluate synthetic speech from 24 TTS systems in 66 languages using transferred decision thresholds.", "boundary": "Spoof rejection rates without target genuine speech do not establish target-language false-alarm rates.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D06" }, { "arxiv": "2507.01750", "version": "2507.01750v1", "title": "Generalizable Detection of Audio Deepfakes", "authors": [ "Jose A. Lopez", "Georg Stemmer", "Héctor Cordourier Maruri" ], "published": "2025-07-02T14:28:11Z", "updated": "2025-07-02T14:28:11Z", "field": "detection", "url": "https://arxiv.org/abs/2507.01750v1", "group": "D2", "label": "Generalizable Detection", "problem": "Identify training choices that improve unseen-condition detection.", "approach": "Compare pretrained speech backbones, augmentation and loss choices across out-of-domain evaluations.", "boundary": "Use held-out generator and dataset tests to distinguish transfer from dataset fit.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D07" }, { "arxiv": "2511.19974", "version": "2511.19974v1", "title": "Continual Audio Deepfake Detection via Universal Adversarial Perturbation", "authors": [ "Wangjie Li", "Lin Li", "Qingyang Hong" ], "published": "2025-11-25T06:41:11Z", "updated": "2025-11-25T06:41:11Z", "field": "detection", "url": "https://arxiv.org/abs/2511.19974v1", "group": "D2", "label": "Continual detection with UAP", "problem": "Learn new attacks without forgetting earlier fake speech.", "approach": "Use universal adversarial perturbations to retain information about past distributions during continual training.", "boundary": "The task sequence and replay assumptions define the strength of the result.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D08" }, { "arxiv": "2601.20510", "version": "2601.20510v1", "title": "Audio Deepfake Detection in the Age of Advanced Text-to-Speech models", "authors": [ "Robin Singh", "Aditya Yogesh Nair", "Fabio Palumbo", "Florian Barbaro", "Anna Dyka", "Lohith Rachakonda" ], "published": "2026-01-28T11:39:40Z", "updated": "2026-01-28T11:39:40Z", "field": "detection", "url": "https://arxiv.org/abs/2601.20510v1", "group": "D2", "label": "Advanced TTS evaluation", "problem": "Check whether newer synthesis families break existing detectors.", "approach": "Generate a controlled corpus with three synthesis systems and compare four detection approaches.", "boundary": "Its multi-view findings concern the tested systems, not all modern TTS.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D09" }, { "arxiv": "2609.19818", "version": "2609.19818v3", "title": "CoReLoop: Parameter-Efficient Controlled Recurrent Refinement for Audio Deepfake Detection", "authors": [ "Kunyu Feng", "Yuxiang Wang", "Li Wang", "Wan Lin", "Zhizheng Wu" ], "published": "2026-09-17T07:28:40Z", "updated": "2026-10-13T11:44:24Z", "field": "detection", "url": "https://arxiv.org/abs/2609.19818v3", "group": "D2", "label": "CoReLoop", "problem": "Improve an existing detector without retraining its whole backbone.", "approach": "Add controlled recurrent refinement through low-rank adaptation and lightweight modules.", "boundary": "Extra inference passes have a compute cost; compare quality at matched deployment budgets.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D10" }, { "arxiv": "2609.37586", "version": "2609.37586v1", "title": "Learning as Deepfakes Evolve: RF-Prompt for Continual Audio Deepfake Detection", "authors": [ "Yuankun Xie", "Xiaoxuan Guo", "Xiaopeng Wang", "Siqing Qin", "Shaole Li", "Kong Aik Lee" ], "published": "2026-09-29T13:43:38Z", "updated": "2026-09-29T13:43:38Z", "field": "detection", "url": "https://arxiv.org/abs/2609.37586v1", "group": "D2", "label": "RF-Prompt", "problem": "Separate learning new generators from learning new real-speech domains.", "approach": "Introduce a controlled incremental protocol and shared real-speech prompts with expanding fake experts.", "boundary": "The protocol controls specific shifts; deployment still needs broader chronological testing.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D11" }, { "arxiv": "2610.05264", "version": "2610.05264v1", "title": "Task-Aware Joint Pruning and Distillation for Efficient Audio Deepfake Detection", "authors": [ "Miao He", "Peng Cheng", "Zhongjie Ba", "Qing Wen", "Li Lu", "Xin Yang", "Kui Ren" ], "published": "2026-10-04T14:34:27Z", "updated": "2026-10-04T14:34:27Z", "field": "detection", "url": "https://arxiv.org/abs/2610.05264v1", "group": "D2", "label": "Task-aware compression", "problem": "Put large pretrained detectors on smaller devices.", "approach": "Jointly prune model structure and distill detection knowledge across domains.", "boundary": "Compression must preserve unseen-condition detection and calibration, not only model size.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D12" }, { "arxiv": "2610.16049", "version": "2610.16049v1", "title": "Adaptive Compression and Targeted Perturbation: A Unified Framework for Generalized Audio Deepfake Detection", "authors": [ "Jianqiao Cui", "Haoxian Zhu", "Hanbiao Meng", "Yanghao Zhang" ], "published": "2026-10-12T01:40:48Z", "updated": "2026-10-12T01:40:48Z", "field": "detection", "url": "https://arxiv.org/abs/2610.16049v1", "group": "D2", "label": "Adaptive compression and perturbation", "problem": "Improve compact representations for varying spoof artifacts.", "approach": "Combine adaptive feature compression with targeted mel-band regularization.", "boundary": "Reported comparisons cover selected datasets and settings; no universal winner follows.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D13" }, { "arxiv": "", "version": "", "title": "PartialEdit: Identifying Partial Deepfakes in the Era of Neural Speech Editing", "authors": [ "You Zhang", "Baotong Tian", "Lin Zhang", "Zhiyao Duan" ], "published": "2025-08", "updated": "", "url": "https://www.isca-archive.org/interspeech_2025/zhang25g_interspeech.html", "field": "detection", "group": "D3", "label": "PartialEdit", "problem": "Detect seamless edits produced by neural speech-editing systems.", "approach": "Build an edited-speech dataset and evaluate detection and temporal localization.", "boundary": "The paper reports failure of detectors transferred from PartialSpoof; realistic editing changes the test distribution.", "evidence": "Official Interspeech 2025 abstract and paper checked.", "venue": "Interspeech 2025", "doi": "", "ref": "D14" }, { "arxiv": "2604.02913", "version": "2604.02913v1", "title": "Split and Conquer Partial Deepfake Speech", "authors": [ "Inbal Rimon", "Oren Gal", "Haim Permuter" ], "published": "2026-04-03T09:33:01Z", "updated": "2026-04-03T09:33:01Z", "field": "detection", "url": "https://arxiv.org/abs/2604.02913v1", "group": "D3", "label": "Split and Conquer", "problem": "Locate short fake regions inside otherwise genuine speech.", "approach": "First detect boundaries, then classify segments; train with multiple segment lengths.", "boundary": "Validate localization and classification separately on edits from unseen systems.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D15" }, { "arxiv": "2605.29531", "version": "2605.29531v3", "title": "Half-Truth Audio Detection and Localisation: A Lightweight Cross-Attentive Architecture and a Cross-Corpus Diagnostic Study", "authors": [ "S. Sutharya", "Remya K. Sasi" ], "published": "2026-05-28T07:47:22Z", "updated": "2026-09-02T12:13:09Z", "field": "detection", "url": "https://arxiv.org/abs/2605.29531v3", "group": "D3", "label": "Half-Truth / CAFNet", "problem": "Detect and localize edits with a small model.", "approach": "Fuse complementary acoustic features using cross-attention and predict fake-region boundaries.", "boundary": "The checked revision reports near-chance binary transfer to PartialSpoof despite stronger in-domain results.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D16" }, { "arxiv": "2606.10912", "version": "2606.10912v1", "title": "What Do Deepfake Speech Detectors Actually Hear?", "authors": [ "Vojtěch Staněk", "Veronika Jirmusová", "Anton Firc", "Kamil Malinka", "Jakub Reš", "Martin Perešíni" ], "published": "2026-06-09T14:21:45Z", "updated": "2026-06-09T14:21:45Z", "field": "detection", "url": "https://arxiv.org/abs/2606.10912v1", "group": "D3", "label": "What Do Detectors Hear?", "problem": "Identify the acoustic evidence behind detector decisions.", "approach": "Use time-aligned attribution, manual cue annotation and causal masking with three WavLM-based detectors.", "boundary": "Attribution explains these models on the tested data; it is not forensic proof for every recording.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D17" }, { "arxiv": "2609.10051", "version": "2609.10051v1", "title": "Zero-Shot Temporal Localisation of Audio Deepfakes in Multi-Speaker Conversations", "authors": [ "Soumyadeep Roy" ], "published": "2026-09-09T11:27:26Z", "updated": "2026-09-09T11:27:26Z", "field": "detection", "url": "https://arxiv.org/abs/2609.10051v1", "group": "D3", "label": "Zero-shot conversation localization", "problem": "Locate inserted synthetic sentences in multi-speaker recordings.", "approach": "Wrap frozen detectors with window scoring and temporal decoding without retraining.", "boundary": "The benchmark uses constructed conversations; natural recordings need separate validation.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D18" }, { "arxiv": "2503.19176", "version": "2503.19176v2", "title": "SoK: How Robust is Audio Watermarking in Generative AI models?", "authors": [ "Yizhu Wen", "Ashwin Innuganti", "Aaron Bien Ramos", "Hanqing Guo", "Qiben Yan" ], "published": "2025-03-24T21:57:59Z", "updated": "2025-03-27T00:51:02Z", "field": "detection", "url": "https://arxiv.org/abs/2503.19176v2", "group": "D4", "label": "Audio watermarking SoK", "problem": "Assess whether embedded provenance survives signal changes.", "approach": "Systematically examine watermark robustness in generative-audio settings.", "boundary": "Adjacent provenance evidence; absence of a detectable watermark is not proof of genuine speech.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D19" }, { "arxiv": "2605.09568", "version": "2605.09568v3", "title": "RADAR Challenge 2026: Robust Audio Deepfake Recognition under Media Transformations", "authors": [ "Hieu-Thi Luong", "Xuechen Liu", "Ivan Kukanov", "Zheng Xin Chai", "Kong Aik Lee" ], "published": "2026-05-10T14:29:35Z", "updated": "2026-05-25T00:50:21Z", "field": "detection", "url": "https://arxiv.org/abs/2605.09568v3", "group": "D4", "label": "RADAR 2026", "problem": "Evaluate detection after ordinary media processing.", "approach": "Benchmark compressed, resampled, noisy and reverberant audio under shared challenge conditions.", "boundary": "Report threshold behavior and false alarms alongside benchmark error rates.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D20" }, { "arxiv": "2605.23201", "version": "2605.23201v2", "title": "MixFake: Benchmarking and Enhancing Audio Deepfake Detection in Diverse Real-world Mixed Audio", "authors": [ "Qingcao Li", "Yipeng Lin", "Weichen Lian", "Zhongjie Ba", "Peng Cheng", "Zhichao Lian" ], "published": "2026-05-22T03:33:36Z", "updated": "2026-10-08T02:12:32Z", "field": "detection", "url": "https://arxiv.org/abs/2605.23201v2", "group": "D4", "label": "MixFake", "problem": "Detect synthetic components in mixed real-world audio.", "approach": "Create mixed-authenticity foreground and background examples; inject signal-level priors through multi-stream prompts.", "boundary": "Mixed audio needs source-aware labels; a single file label can hide the target task.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D21" }, { "arxiv": "2606.29544", "version": "2606.29544v1", "title": "Proteus: Automated Adversarial Robustness Testing for Audio Deepfake Detectors", "authors": [ "Nicolas M. Müller", "Aditya Tirumala Bukkapatnam", "Zohaib Ahmed" ], "published": "2026-06-28T18:14:28Z", "updated": "2026-06-28T18:14:28Z", "field": "detection", "url": "https://arxiv.org/abs/2606.29544v1", "group": "D4", "label": "Proteus", "problem": "Find failures caused by combinations of common transformations.", "approach": "Automate robustness testing against a production detector and use failures for targeted retraining.", "boundary": "Evidence comes from a specific detector; other systems need independent audits.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D22" }, { "arxiv": "2608.05507", "version": "2608.05507v1", "title": "AffectDF: The Most Comprehensive Benchmark for Speech Deepfake Detection against Emotionally Expressive Attacks", "authors": [ "Aurosweta Mahapatra", "Xiutian Zhao", "Shreeram Suresh Chandra", "Zihan Zhang", "Zongyang Du", "Ismail Rasim Ulgen", "Kong Aik Lee", "Nicholas Andrews", "Carlos Busso", "Berrak Sisman" ], "published": "2026-08-06T01:19:35Z", "updated": "2026-08-06T01:19:35Z", "field": "detection", "url": "https://arxiv.org/abs/2608.05507v1", "group": "D4", "label": "AffectDF", "problem": "Test detectors against expressive rather than neutral fake voices.", "approach": "Evaluate emotional TTS, conversion and audio-language-model attacks across emotion conditions.", "boundary": "Adding emotional training data does not consistently solve transfer to unseen attacks.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D23" }, { "arxiv": "2608.23437", "version": "2608.23437v2", "title": "AT-ADD: A Benchmark and Challenge for Robust and All-Type Audio Deepfake Detection", "authors": [ "Yuankun Xie", "Haonan Cheng", "Jiayi Zhou", "Xiaoxuan Guo", "Tao Wang", "Changhao Zhang", "Jian Liu", "Weiqiang Wang", "Ruibo Fu", "Xiaopeng Wang", "Hengyan Huang", "Xiaoying Huang", "Long Ye", "Guangtao Zhai", "Kong Aik Lee" ], "published": "2026-08-24T16:11:43Z", "updated": "2026-10-15T12:47:30Z", "field": "detection", "url": "https://arxiv.org/abs/2608.23437v2", "group": "D4", "label": "AT-ADD", "problem": "Measure robustness to unseen generators, recording channels and replay.", "approach": "Run robust-speech and all-audio tracks with shared baselines and submitted systems.", "boundary": "This map uses the speech track; the all-audio track also includes music, singing and environmental sounds.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D24" }, { "arxiv": "2608.24159", "version": "2608.24159v1", "title": "On the Robustness of Audio Deepfake Detection under Audio Watermarking", "authors": [ "Zi Qian Yong", "Ajinkya Kulkarni", "Julia Lau", "Hwa Hui Tew", "Shu Min Leong", "Raphael Phan", "Sébastien Marcel" ], "published": "2026-08-25T07:18:16Z", "updated": "2026-08-25T07:18:16Z", "field": "detection", "url": "https://arxiv.org/abs/2608.24159v1", "group": "D4", "label": "Watermark impact on detectors", "problem": "Check whether benign provenance marks disturb detection.", "approach": "Measure detector scores and representation shifts after watermarking across several datasets.", "boundary": "The impact is dataset-dependent; watermarking and blind detection need joint testing.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D25" }, { "arxiv": "2609.21738", "version": "2609.21738v1", "title": "GenTraceBench: A Benchmark for Tracing Audio Deepfakes Across Pre- and Post-training Stages", "authors": [ "Li Wang", "Kunyu Feng", "Wan Lin", "Dekun Chen", "Qinke Ni", "Xueyao Zhang", "Lei Wang", "Jie Shi", "Haizhou Li", "Zhizheng Wu" ], "published": "2026-09-18T13:15:52Z", "updated": "2026-09-18T13:15:52Z", "field": "detection", "url": "https://arxiv.org/abs/2609.21738v1", "group": "D4", "label": "GenTraceBench", "problem": "Test whether generator fingerprints survive model adaptation.", "approach": "Hold texts and prompts fixed across pretraining and post-training variants; test detection and source tracing.", "boundary": "Fingerprint stability differs by adaptation method and forensic backbone.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D26" }, { "arxiv": "2610.10752", "version": "2610.10752v1", "title": "Detection-Guided Adaptive Purification with Diffusion Models for Robust Audio Deepfake Detection", "authors": [ "Muhammed Salih Kayhan", "Qiben Yan" ], "published": "2026-10-07T18:19:55Z", "updated": "2026-10-07T18:19:55Z", "field": "detection", "url": "https://arxiv.org/abs/2610.10752v1", "group": "D4", "label": "Adaptive purification", "problem": "Preserve detection evidence while removing adversarial perturbations.", "approach": "Adjust diffusion purification strength per input using detector score changes.", "boundary": "The tested attack and detector settings define the defense claim; benign-audio quality also matters.", "evidence": "Primary abstract checked; editorial interpretation marked by evidence boundary.", "ref": "D27" } ]