https://arxiv.org/api/dN3Wjd/+PGnjvv2KZWnJhWuv8KI 2026-09-21T19:22:32Z 30 2461 0 http://arxiv.org/abs/2609.21738v1 2026-09-18T13:15:52Z

Modern text-to-speech (TTS) systems are rarely deployed as unchanged pre-trained models. They are often adapted through supervised fine-tuning (SFT) or preference optimization such as DPO and GRPO. This raises a practical question for audio deepfake forensics: do fingerprints learned from a foundation generator remain valid after adaptation? We present GenTraceBench, a controlled benchmark spanning five TTS architectures, 16 pre-/post-training variants, and 49,728 utterances generated with fixed texts and speaker prompts. Under a train-on-foundation, test-on-adapted protocol, we evaluate binary detection, closed-set attribution, and open-set verification. DPO and GRPO generally preserve fingerprints, whereas some SFT and pre-training-data changes cause substantial drift; effect sizes vary across three forensic backbones. Repeated training runs confirm the largest W2V-BERT attribution drop, while a data-mixture control with comparable speech quality shows that composition change need not cause drift. In W2V-BERT verification, multi-shot enrollment reduces EER for the SFT condition from 44.4% to 11.0%, whereas the SingNet-only condition remains at or above 45% EER.

2026-09-18T13:15:52Z 5 pages, 2 figures, 4 tables. Accepted to the 15th International Symposium on Chinese Spoken Language Processing (ISCSLP 2026) Li Wang Kunyu Feng Wan Lin Dekun Chen Qinke Ni Xueyao Zhang Lei Wang Jie Shi Haizhou Li Zhizheng Wu http://arxiv.org/abs/2609.21581v1 2026-09-18T10:11:58Z

Audio deepfake forensics is moving beyond a simple real-or-fake verdict toward attribution: which system generated the audio clip? Most source-attribution methods cast this as closed-set classification, so they cannot name a generator that was absent from training, a gap that widens with every newly released text-to-speech (TTS) system. We instead frame attribution as cross-modal retrieval: each generator is described in natural language, and a clip is attributed by retrieving the description closest to it in a shared audio-text embedding space. Adding a new system then takes nothing more than writing its description, with no retraining and no new classifier head. Our model couples a frozen Wav2Vec2-BERT audio encoder with a frozen E5 text encoder and aligns them through small trainable projection heads, using a contrastive objective that combines a cross-modal supervised-contrastive loss with intra-modal terms. We evaluate on MLAAD v9 (140 TTS models, 51 languages) under 10-fold leave-models-out cross-validation. For generators it has never encountered before, the model reaches a model-level mean reciprocal rank (MRR) of 58.4%. Even when the correct model is not identified, the audio clip is often matched to systems that share the true generator's vocoder, acoustic model, or architecture. Because the same embedding space also answers natural-language attribute queries, one set of descriptions covers both open-set attribution and attribute-level forensic profiling.

2026-09-18T10:11:58Z Accepted to the 2026 IEEE Spoken Language Technology Workshop (SLT 2026) Cristian-Teodor Neamtu Serban Mihalache Stefan Smeu Dan Oneata Horia Cucu Dragos Burileanu http://arxiv.org/abs/2609.19805v1 2026-09-17T07:16:45Z

Grapheme-to-phoneme (G2P) conversion turns raw text into its phonemic form and is an essential part of both text-to-speech (TTS) and automatic speech recognition (ASR) systems. It is required to be fast, stable and context-aware. For unsegmented languages such as Japanese, G2P additionally couples word segmentation with highly context-dependent polyphone disambiguation, and the scarcity of accurately annotated data remains a bottleneck. In this paper, we present a context-aware neural G2P method that scores paths of a discriminative conditional random field (CRF) over a word lattice constructed from dictionaries. To tackle data scarcity, we utilize large language models (LLMs) to generate more than 2 million sentences. Experimental results demonstrate that our method strongly outperforms conventional morphological analyzer-based methods and neural sequence models. On the Joyo-Kanji-Yomi benchmark, our method reaches 99.62% target word reading accuracy, 0.32% target word phoneme error rate (PER) and 0.14% sentence PER.

2026-09-17T07:16:45Z Submitted to ICASSP 2027 Rui Hu Zhenpeng Zhan Xiaolong Lin http://arxiv.org/abs/2609.19324v1 2026-09-16T18:45:01Z

Advancement of zero-shot text-to-speech synthesis is currently hindered for low-resource languages by the scarcity of large-scale, high-fidelity speech datasets. Traditional alignment-based methods require rare verbatim transcripts, while standard in-the-wild pipelines often rely on single-model automatic speech recognition and silence-based segmentation, leading to transcription errors and truncated prosody. To address these challenges for the Persian language, this paper introduces PersianVox, a fully automated pipeline designed to generate high-quality speech corpora from unlabeled web data. Our approach integrates a novel prosody-aware segmentation strategy that utilizes acoustic turn-detection to preserve linguistic completeness and optimize utterance duration for long-context modeling. Furthermore, we employ a dual-model agreement mechanism, leveraging two distinct model architectures to filter unreliable transcriptions without ground truth. This pipeline yields a 2,400-hour multi-speaker dataset, the largest open-source speech resource available for Persian to date. Additionally, we provide the first comparative benchmark of speech quality assessment methods for Persian, releasing a human-annotated subset to facilitate future research.

2026-09-16T18:45:01Z 4 pages, 3 figures, 5 tables Saeedreza Zouashkiani Soheil Khalesi Saman Soleimani Roudi Sajjad Amini Shahrokh Ghaemmaghami http://arxiv.org/abs/2609.17509v1 2026-09-15T17:46:51Z

Neural audio codecs are a key component in speech language modeling. However, their high frame rates lead to long sequence lengths, increasing computational costs. Dynamic frame rate codecs mitigate this by reducing the effective frame rate using a compression step to merge multiple frames together. However, most prior methods either operate on single-codebook codecs or apply a single compression step before multi-layer quantization. This forces all quantization layers to share the same segmentation boundaries, despite the residual embeddings at different quantization layers exhibiting different rates of change over time. We propose LACE (Layer-Adaptive Codec Encoding), a dynamic frame rate codec that applies an independent compression step at each quantization layer, enabling layer-specific segmentation boundaries. To use LACE tokens in downstream text-to-speech (TTS), we further introduce union alignment and boundary anchor mechanisms to make durations consistent across layers while preserving compression benefits. Experiments on LibriTTS show that LACE offers a better rate-quality tradeoff than prior dynamic frame rate methods on the reconstruction task and improves TTS inference efficiency while maintaining competitive synthesis quality. Our code is released as part of the ESPnet3 codec recipe.

2026-09-15T17:46:51Z Accepted to SLT 2026. 8 pages, 5 figures Thanapat Trachu Samuele Cornell William Chen Shinji Watanabe http://arxiv.org/abs/2609.17234v2 2026-09-16T02:54:01Z

Text-to-speech that reads raw text has no lexicon: a rare word is read as guessed. Remedies train a reading-and-accent channel on recorded speech or edit words one at a time from exemplars. We do neither. The frozen backbone reads a sentence containing a common word it already says correctly, and its own output then serves as the teacher for the same sentence, with that word replaced by a tagged, accented reading; this training pair is the whole idea. On Sarashina2.2-TTS, screened raters at Fleiss' kappa = 0.85 hear the prescribed accent on 0.89 of unseen words against 0.57 for kana, which cannot express one; kana wins no pair; naturalness is not measurably hurt. Moved untuned to autoregressive, diffusion, and encoder-decoder backbones, it transfers reading, 0.25 to 0.47 above no edit on 319 words, and on CosyVoice 2 accent on two words in three, but not on Irodori; the paper locates why.

2026-09-15T14:13:51Z Since the ICASSP deadline was extended, I decided to redo the experiment Shuhei Kato http://arxiv.org/abs/2609.16989v1 2026-09-15T11:03:52Z

Long-form text-to-speech (TTS) enables multi-turn conversations with consistent prosody and higher quality voice cloning from longer reference audio. Recent open-weights autoregressive TTS models such as Qwen3-TTS and VoxCPM2 attain state-of-the-art word error rate (WER) and speaker similarity (SIM) on short-form prompts but significantly deteriorate when used with long-form prompts. We propose Localized Attention-Constrained Inference (LACI), an inference-only method to detect TTS errors in near real-time, roll back to the error onset and regenerate with temporary guardrails, adding negligible computational overhead. Using LACI, we improve worst-of-N WER across 10 RNG seeds for Qwen3-TTS-0.6B from 35.2% to 3.4% on prompts longer than 1500 words, even surpassing its short-form reliability of 5.4\% on prompts with fewer than 500 words. To demonstrate the efficacy of LACI on voice cloning reliability, we propose a sliding-window version of the SIM metric that we call wSIM. wSIM exposes several novel failure patterns that are not captured by SIM. LACI improves worst-of-N wSIM from 0.01 to 0.47 on 120 seconds of reference audio while reducing the rate of catastrophic generations with WER above 30% from 26% to below 1%

2026-09-15T11:03:52Z Rongxiang Wang Berkin Durmus Aysegul Orhon Eduardo Pacheco Atila Orhon http://arxiv.org/abs/2609.15313v1 2026-09-14T10:03:49Z

Autoregressive generation of interleaved text and acoustic tokens is a common approach to spoken-response generation in speech large language models. Although this design enables streaming generation with explicit textual guidance, generated acoustic tokens become part of the context for subsequent text predictions. Given identical speech inputs, we observe markedly lower answer accuracy for the internal text generated in speech-to-text-and-speech (S2TS) mode than for speech-to-text (S2T) responses. We term this discrepancy the \emph{output-mode gap} (OMG). To reduce OMG, we propose \emph{Joint-Output On-Policy Distillation} (JO-OPD), which distills the model's stronger S2T policy into joint generation using student-generated S2TS trajectories. At each text position, the S2T teacher provides soft targets from a text-only projection of the student's preceding outputs, while the student predicts from the corresponding full interleaved history. A preservation objective further regularizes native non-text predictions. Experiments on Step-Audio-2-mini and Baichuan-Audio-Instruct reveal OMG across two interleaved generation architectures. On Step-Audio-2-mini, JO-OPD reduces OMG from 42.87 to 16.26 percentage points on Spoken-MQA and from 29.72 to 13.04 points on speech-rendered GSM8K, with little change in S2T accuracy and substantially larger reductions than matched SFT baselines. ASR-based evaluation further shows a 7.49-point improvement in spoken-answer accuracy on Spoken-MQA.

2026-09-14T10:03:49Z Daxin Tan Dehua Tao Chengxi Deng Hanlin Zhang Xiao Chen http://arxiv.org/abs/2609.15184v1 2026-09-14T08:05:26Z

Zero-shot text-to-speech (TTS) can clone a speaker's voice from a short audio prompt, yet most TTS systems still require the audio prompt transcript during inference. This dependency prevents cross-lingual voice cloning when the audio prompt transcript is unavailable, particularly for unseen languages. Cross-Lingual F5-TTS removes this dependency and enables transcript-free cross-lingual voice cloning, but it prepares its training data with forced alignment. Forced alignment is sensitive to boundary errors, and its cost grows as more languages are covered. Its speaking rate predictor is also unreliable at estimating duration when the audio prompt begins or ends with silence. In this paper, we present Cross-Lingual F5-TTS 2, a simplified framework for transcript-free cross-lingual voice cloning without forced alignment. Instead of using forced alignment to segment real utterances, we build same-speaker prompt and target pairs using a pretrained F5-TTS model and fine-tune the same model on these constructed pairs. This simplifies data preparation and preserves the acoustic modeling capability of the pretrained model, enabling adaptation with only a short fine-tuning stage. We further make the syllable-level speaking rate predictor robust to leading and trailing silence through silence-aware augmentation. Experiments show that Cross-Lingual F5-TTS 2 reaches higher speaker similarity than F5-TTS and Cross-Lingual F5-TTS while maintaining intelligibility. All related resources are publicly available.

2026-09-14T08:05:26Z Qingyu Liu Rixi Xu Yushen Chen Zhikang Niu Haitao Li Pengcheng Zhu Bowen Zhang Jian Zhao Yunting Yang Qinyuan Cheng Xipeng Qiu Berrak Sisman Kai Yu Xie Chen http://arxiv.org/abs/2609.14817v1 2026-09-13T22:18:35Z

In Yorùbá, pitch alone separates \d{o}k\d{o} (husband, Mid), \d{o}k\d{ò} (vehicle, Low), and \d{o}k\d{ó} (hoe, High) -- the diacritics ARE the tone marks. Yet character error rate (CER), the standard automated metric for text-to-speech (TTS), is in practice computed from ASR output that drops those marks: a synthesizer can ace CER and still say vehicle for husband. We introduce DunDun -- named for the dùndún, the Yorùbá talking drum that speaks through pitch alone -- an automated, reference-free lexical-tone metric that needs no tone-labelled corpus. The gold High/Mid/Low sequence is read from the input text's diacritics (in TTS that text exists by construction, so no reference recording is needed); the prediction comes from the audio's pitch track. We validate three ways. Flattening pitch with PSOLA resynthesis collapses DunDun while CER does not move. Inverting High and Low in the answer key of 300 native recordings drives the two-class readout to 0.14, symmetrically below its 0.35 chance level -- a consistency check on the scoring path, not independent evidence. And three native listeners, over 67 blind A/B trials, pick the tone-correct clip 89.6% of the time (95% CI 80.0-94.8; p < 1e-4); whether DunDun tracks those judgements trial by trial is not resolved at this sample size. Applied to a massively multilingual zero-shot TTS model, DunDun shows what CER cannot: Yorùbá tone sits near the native anchor before any Yorùbá fine-tuning (0.567 +/- 0.02 over five decode seeds vs. 0.596; chance 0.33), despite the 21.4% CER the model's own paper reports; and a few hours of clean audio halve CER (5.6% to 2.7% by 5h, 1.7% by 15h) while tone saturates within the hour. On non-tonal Swahili, CER already captures the gains: the metric a language needs is language-dependent. We release the metric and the complete validation protocol.

2026-09-13T22:18:35Z 9 pages, 2 figures Moses Daudu Adeola Enitan Bamidele Honor-Jesus Bezaleel http://arxiv.org/abs/2609.14743v1 2026-09-13T19:09:07Z

Pure speech language models often lag behind text and speech-text language models in generating coherent content, but this gap is difficult to quantify because speech and text systems are typically evaluated with different metrics and trained on different data. We study the speech-text modality gap in a family of spoken language models, based on flow matching for continuous acoustic feature generation. We construct a unified generation-based evaluation suite that compares speech-only, text-only, and speech-text language models trained on matched data distributions and evaluated in matched generation settings. We evaluate generated continuations along multiple dimensions: semantic coherence, measured by transcribing generated speech and scoring it with a reference language model; local phonetic structure, measured by phone n-gram distributional statistics; speaker consistency and acoustic quality; and emotion-based distributional metrics. Across datasets, we find that joint speech-text modeling substantially improves semantic coherence. However, the improvement is not uniform across metrics: phone-level metrics change only modestly, speaker similarity and predicted quality are lower for speech-text continuations, while emotion-based distributional metrics improve. Compared with larger-scale speech-only models, our speech-text model closes much of the scaling gap in transcript-based semantic coherence, suggesting that text provides an efficient semantic training signal for spoken language modeling.

2026-09-13T19:09:07Z Accepted to SLT 2026, extended version with appendix Ju-Chieh Chou Jiawei Zhou Karen Livescu http://arxiv.org/abs/2609.14408v1 2026-09-13T10:13:10Z

We present an automated pipeline that converts NCERT textbooks into interactive video explanations that respond directly to user queries. A user uploads a PDF and asks a question; the system then generates a video-based explanation as output, handling both text and visual elements from the PDF for multi-modal retrieval and response generation. The pipeline combines a Retrieval-Augmented Generation (RAG) model with generative multimedia components. The RAG stage is optimized for the structure of NCERT textbooks and performs best on content from those books. Given a user query, the RAG model retrieves relevant content from the PDF and generates a multi-scene script containing narrative explanations and structured visual prompts aligned with the textbook's explanatory style. These prompts are passed to a Stable Diffusion module, implemented layer by layer for interpretability and control, which generates contextually relevant images. The images are then processed by DynamiCrafter to produce animated sequences. Finally, a Google Text-to-Speech module generates synchronized narration, aligning speech with the visual scenes through time-based control. The result is a coherent video explanation integrating animation, narration, and textbook-aligned visuals, transforming static educational material into an engaging learning experience. By combining multi-modal document retrieval, generative visual models, animation frameworks, and speech synthesis, this pipeline demonstrates a scalable approach to delivering interactive, personalized digital education content.

2026-09-13T10:13:10Z Siddhanth Sridhar Shreya Chaurasia Baddela Sai Yaswantha Reddy Deepak Parmar Shylaja S S http://arxiv.org/abs/2609.13869v1 2026-09-12T10:45:24Z

Automatic sentence function identification is important for many downstream natural language processing (NLP) applications such as dialogue systems, text-to-speech synthesis, and machine translation. However, benchmark resources for Bangla sentence function classification remain limited. To mitigate this gap, this paper introduces a corpus of 10,000 Bangla sentences, manually annotated into four functional categories, namely declarative, interrogative, imperative, and exclamatory. The corpus is nearly balanced across the four classes, with high annotation reliability reflected by a Fleiss\' Kappa of 0.82. Furthermore, we evaluate multiple feature representations, including Bag-of-Words (BoW), TF-IDF, and Word2Vec, with several classical machine learning classifiers. In addition, two heterogeneous ensemble models, namely Single-Level Ensemble (SLE) and Double-Level Ensemble (DLE), are utilized to improve classification performance. Experimental results show that TF-IDF consistently outperforms Word2Vec, likely due to its ability to emphasize discriminative lexical cues associated with sentence functions, particularly given the relatively small corpus used to train Word2Vec. The DLE model with TF-IDF features achieves the best performance with accuracy and macro-F1 of 0.95, demonstrating the effectiveness of sparse lexical representations and heterogeneous ensemble learning for this task. Further cross-validation confirms the robustness of the approach, while LIME-based interpretability provides insights into model predictions. The developed corpus and model benchmarking establish strong baselines for Bangla sentence function classification.

2026-09-12T10:45:24Z Accepted at 2026 IEEE 5th International Conference on Robotics, Automation, Artificial-Intelligence and Internet-of-Things (RAAICON) Swapnil Kundu Argha Abdullah Al Shafi Rowzatul Zannat Shoumik Barman Polok Abdul Muntakim Jannatul Ferdousi M. A. Moyeen http://arxiv.org/abs/2609.13792v1 2026-09-12T08:11:02Z

We present the results of the VoiceMOS Challenge 2026, the fifth edition of a scientific challenge on automatic prediction of subjective speech assessments. After expanding the scope to music and general audio in 2025, we refocused the evaluation target on speech and organized three tracks: prediction of absolute and comparative category ratings for enhanced speech, prediction of naturalness and emotion-related tasks for emotional text-to-speech systems, and prediction of speaker and accent similarity for codec-based speech synthesis systems. The challenge attracted a total of 18 teams worldwide, with most teams successfully surpassing the provided baselines. We summarize the challenge results, representative top-performing systems, participant feedback, and directions for future editions.

2026-09-12T08:11:02Z Preprint Wen-Chin Huang Wei Wang Marvin Sach Xiaoxue Gao Nicholas Sanders Erica Cooper Toda Tomoki http://arxiv.org/abs/2609.12945v1 2026-09-11T15:05:20Z

We introduce StepAudio 3 Gen, a general-purpose audio generation model that supports zero-shot text-to-speech (TTS), voice design, vocal generation, sound effects, music, vibe speech, and mixtures of multiple audio types within a unified framework. At its core, StepAudio 3 Gen is a discrete autoregressive generator that models audio directly over residual vector quantization (RVQ) tokens, departing from the diffusion Transformer-based continuous generation paradigm prevalent in recent general audio models. Its StepAudio Tokenizer represents general audio at 12.5 Hz in a shared $16 \times 2048$ residual code space, jointly quantizing semantic and waveform-level acoustic features so that each code layer preserves both types of information. For generation, the backbone predicts the first codebook along the time axis using autoregressive modeling, while a lightweight causal Transformer completes the remaining fifteen codebooks along the codebook axis. Our study further identifies three key design principles: (1) interference-aware progressive pretraining for acquiring audio capabilities while preserving the textual abilities of the large language model, (2) RVQ Adaptor for effectively incorporating multi-codebook acoustic representations, and (3) discrete autoregressive modeling over a shared representation across general audio domains. With progressive pretraining, multi-task instruction training, and supervised fine-tuning, StepAudio 3 Gen achieves state-of-the-art performance on both TTS and voice design, while retaining strong generation capabilities across speech, vocals, sound effects, and music. Audio samples are available at https://stepaudiollm.github.io/step-audio-3-gen/.

2026-09-11T15:05:20Z Bin Lin Bo Zhao Boyang Wang Boyang Zhang Boyong Wu Chao Yan Chen Geng Chen Wu Cheng Yi Chengli Feng Chenglin Zhu DanNi Wan Daxin Jiang Dongqing Pang Fei Tian Feng Tian Future Li Gang Yu Guanglong Yang Jia Peng Jiahao Song Jiamin Fan Jiangjie Zhen Jianzheng Gao Jun Chen Li Xie Lifang Zhang Lingli Ji Liying Shi Lun Cai Min Xu Na Wang Peilin Li Peng Yang Pengfei Tan Qingjian Lin Ruijie Xiong Runze Li Shenghua Hu Shi Qiu Siqi Tu Siyi Zhou Tianjiao Deng Wanying Lu Weiming Niu Wen Sun WenWen Qu Xiangyu Zhang Xianwei Zhang XiaoSu Su Xing Chen Xinyu Liu Xuerui Yang Yang Li Yang Yang Yechang Huang Yibo Zhu Yifan Zhang Yiyang Xu Yu Fu Yu Luo Yu Zhou Yumang Wang Yunzhou Ju Yuxiang Yang Zekai Liu Zengwei Yao Zhenwei Mou Zheqi Dai Zhiyue Wu Zichao Zhou http://arxiv.org/abs/2609.12918v1 2026-09-11T14:44:01Z

A vocoder is a pivotal component of modern text-to-speech (TTS) systems. Despite the significant progress of neural network-based vocoders, accurate phase reconstruction remains the main challenge limiting both audio quality and modeling efficiency. We introduce PhaseGAN, a lightweight vocoder that addresses this limitation through a "mel $\rightarrow$ Amplitude $\rightarrow$ Phase" reconstruction pipeline. By reconstructing amplitude and phase spectra via distinct methodologies, the proposed PhaseGAN outperforms state-of-the-art baselines while utilizing fewer model parameters and reduced computational requirements. The compact version generates high-fidelity audio with approximately 500K parameters and 1 GMAC computational load, making it highly suitable for real-time applications on edge devices. In addition, our approach exhibits exceptional musical audio synthesis capabilities despite no training on musical data, illustrating unprecedented cross-domain generalization. See https://github.com/phasegan/phasegan-audio-demo for demos of our work.

2026-09-11T14:44:01Z 15 pages, 1 figure Wenzheng Zhang Xueliang Zhang Shulin He Fei Zhao Xin Liu Pengjie Shen Zhenlong Guo Zixuan Xue Hongtao Bao Zixuan Li http://arxiv.org/abs/2609.12855v1 2026-09-11T13:45:06Z

Decoder-only text-to-speech (TTS) models scale efficiently but remain prone to content hallucinations that arise from weak text-speech alignment during autoregressive generation. We find that robustness is governed by a non-monotone relation to the sharpness of the alignment-bearing attention heads: a moderate degree is best, whereas over-sharpening is no better than the unaligned backbone and even less robust. Guided by this, we present AlignDPO, a post-training method that reaches this moderate regime by folding a lightweight connectionist-temporal-classification (CTC) alignment term into Direct Preference Optimization (DPO), applied only to the chosen samples, with no architectural or inference-time change. On the Seed-TTS-Eval English set, this significantly reduces the content-hallucination and word error rates relative to a strong DPO baseline and lowers the severe content-hallucination rate to ~0.6% (from 4.4%); a listening study further finds it preferred for naturalness over both the backbone and that baseline. Alignment is thus best learned and kept moderate rather than maximized or imposed at decoding. Audio samples are available at https://align-dpo-demo.vercel.app.

2026-09-11T13:45:06Z 6 pages, 4 figures, 4 tables. Accepted to IEEE SLT 2026 Xiao Zhou Oisín Turbitt Kit Bower-Morris Jonathan Carlton Jamie Stacey Kris Y. Hong http://arxiv.org/abs/2609.12432v1 2026-09-11T04:42:12Z

Large speech corpora support research, but recordings can expose speaker identity because voice remains a recognizable biometric. Meanwhile, speech data derived from media can be difficult to redistribute reliably. We present \emph{VoxTubeS}, a family of speaker-anonymized synthetic speech corpora designed for redistribution, comprising three method families and seven variants derived from the VoxTube corpus, which is distributed under CC BY-NC-SA 4.0, using 1.29M quality-filtered English utterances from 1,511 speakers. The synthesis methods span voice conversion, latent-space anonymization, and controllable text-to-speech. We evaluate VoxTubeS using utterance-level unlinkability, conversation-level linkability and singling-out, downstream speaker verification, linguistic consistency, speaker diversity, and fairness metrics for gender and accents. Our comprehensive analysis exposes a complex trade-off: stronger identity suppression often reduces linkability but sacrifices utility and population diversity, whereas speaker consistency training improves both utterance- and conversation-level privacy while retaining comparable utility and a broader speaker space. Fairness varies independently of aggregate performance. No method dominates; VoxTubeS therefore treats corpus construction as a choice among operating points that balances privacy, utility, diversity, fairness, and responsible redistribution under the source license.

2026-09-11T04:42:12Z Zhe Zhang Yexin Lu Junichi Yamagishi http://arxiv.org/abs/2609.11763v1 2026-09-10T16:17:19Z

Speech deepfake detection (SDD) models are trained on multi-attack datasets containing diverse spoofing systems, such as text-to-speech (TTS) and voice conversion (VC). In standard classifier training on multi-attack datasets, all attacks are treated as one spoofed class, and performance is reported using overall Equal Error Rate (EER). This aggregate view obscures how individual attacks shape learning and generalization. To better understand this attack-level behavior, we first balance TTS and VC exposure using sample and attack omission. We then measure attack-wise EER at inference and analyze attack-wise training loss and predictive entropy to characterize optimization. Results show that attacks contribute unequally: some attacks have high EER sensitivity and concentrated entropy with low loss, indicating strong influence on the decision boundary. We define these as high-impact attacks. To reduce uneven generalization across attacks, we propose a replay-regularized, attack-aware curriculum that steps exposure based on measured attack influence. Experiments on ASVspoof 2019, 2021, ASVspoof 5, and Fake-or-Real show improved overall robustness and reduced attack-level imbalance compared with standard multi-attack training.

2026-09-10T16:17:19Z Accepted to IEEE Spoken Language Technology (SLT) 2026 Avantika Singh Aurosweta Mahapatra Ismail Rasim Ulgen Nicholas Andrews Kong Aik Lee Berrak Sisman http://arxiv.org/abs/2609.11725v1 2026-09-10T15:41:32Z

Text-to-speech (TTS) models commonly address text--speech alignment by expanding phone-level encoder states to frame-level decoder inputs using predicted durations. While this length-regulation step resolves alignment structurally, this use of duration typically changes only where and how often latent states appear, not the values of the states themselves. This paper proposes a continuous-time mechanism for duration-aware acoustic modelling in TTS using neural controlled differential equations (CDEs). We formulate the phone representation as a temporally parameterised control path and use a neural acoustic vector field to produce a continuous-time hidden state whose values evolve with phonetic content and duration-derived timing. The resulting trajectory can be sampled at discrete points and integrated into a standard acoustic decoder pipeline. Objective results contrast CDEs and typical recurrent models. Subjective results suggest that CDE-based models evaluating one phone per step can improve rank-order agreement between synthesised and reference emotion intensity while maintaining comparable emotion-expression quality to a strong baseline. Additional experiments with half-phone step-sizes suggest that temporal resolution changes the trade-off between style tracking and absolute calibration. These results position CDEs as a promising design space for continuous-time and duration-aware style-sensitive TTS.

2026-09-10T15:41:32Z Accepted to IEEE Spoken Language Technology Workshop (SLT) 2026 Mattias Cross Minghui Zhao Anton Ragni http://arxiv.org/abs/2609.11545v1 2026-09-10T13:41:22Z

Low-resource multilingual text-to-speech (TTS) systems have expanded language coverage, but their robustness under complex text inputs remains insufficiently diagnosed. Existing evaluations mainly focus on naturalness, speaker similarity, and content consistency using regular test sentences, while providing limited insight into how multilingual TTS systems fail when handling challenging inputs such as numbers, dates, named entities, long sentences, code-switched expressions, and punctuation-related structures. This paper proposes a complex-text robustness diagnosis framework for low-resource multilingual TTS. We evaluate robustness from three dimensions: content consistency, language consistency, and generation stability. A multilingual robustness testing scheme is designed for Thai, Vietnamese, Swahili, and Indonesian, covering ordinary sentences and multiple types of complex text inputs. We further introduce automatic diagnostic metrics, including character error rate, language identification accuracy, and duration abnormal rate. To support input-level risk analysis before speech generation, we propose a lightweight Text Risk Score (TRS), which estimates synthesis risk from interpretable text features without manual annotation or model training. Experiments on three representative multilingual TTS systems, including OmniVoice, VoxCPM2, and MMS-TTS, show that complex text inputs expose systematic failure patterns that are not fully reflected by ordinary short-sentence evaluation. Different systems exhibit distinct vulnerabilities in number normalization, named entity handling, long-text generation, and code-switched input processing. Furthermore, TRS shows a positive correlation with content errors and duration abnormalities, demonstrating its usefulness as a low-cost pre-synthesis indicator for complex-text risk diagnosis in low-resource multilingual TTS.

2026-09-10T13:41:22Z NCMMSC 2026 accepted Tianlun Zuo Ziyu Zhang Tingzhi Mao Zhonghua Fu Lei Xie http://arxiv.org/abs/2609.11523v1 2026-09-10T13:24:06Z

Audiobook narration, conversational agents, and audiovisual dubbing require speech that conveys changing emotions and adapts its pacing within a single utterance. But most existing TTS systems typically rely on utterance-level style conditioning, making such fine-grained control difficult to achieve. In light of this, and inspired by the success of post-training in large language models, we propose a unified post-training framework that equips pretrained text-to-speech models with natural-language control over segment-level emotion and duration. Supervised fine-tuning establishes instruction-conditioned speech generation, while reinforcement learning with group relative policy optimization refines control accuracy using emotion and duration rewards alongside content and speaker preservation objectives. By reusing the pretrained architecture, our approach avoids additional inference-time control modules. Experiments demonstrate significantly improved fine-grained controllability while maintaining speech intelligibility and speaker identity, highlighting post-training as a practical approach to extending existing speech synthesis models.

2026-09-10T13:24:06Z Lianru Gao Yujie Guo Yong Qin http://arxiv.org/abs/2609.11458v1 2026-09-10T12:32:13Z

Determining the differences between two speakers' accents is a fundamental task in linguistics and speech technology research. The methodology used to measure these differences depends on the specific research area. A phonetics researcher may demonstrate accent variation by comparing vowel formants in paired recordings of individual words. These results will be interpretable, but the recordings will be time-consuming to collect and may not be representative of connected speech. Accented Text-to-Speech (TTS) research has pushed towards using accent embeddings derived from accent classification tasks. These embeddings can be produced from any speech recording, but are not readily interpretable. In this paper, we demonstrate that articulatory representations created through articulatory inversion can be used as an interpretable basis for accent comparison and that optimal transport provides a framework for accent comparison across arbitrary recording types.

2026-09-10T12:32:13Z Charles McGhee Mark J. F. Gales Kate M. Knill http://arxiv.org/abs/2609.10199v1 2026-09-09T14:04:39Z

We describe a participatory design approach toward developing Deaf-centric text-to-speech (TTS) technologies. While TTS is growing rapidly in the mainstream, it has received little attention to date in the deaf and hard of hearing (DHH) technology space. Critical problems have remained unaddressed for DHH users, including the ability to manipulate tone, emotions and delivery via non-auditory means. Verifying that the generated speech matches intent and is appropriate for a given situation without having to listen to it is another challenge. Respecting cultural and identity factors in the generated speech is also important. This work explores the design space with DHH participants through two focus groups, three co-design sessions, and four one-on-one early-stage design evaluation sessions. Participants included people both familiar and unfamiliar with TTS, as well as DHH content creators. We describe key findings, design ideas, results, and implications for future Deaf-centric TTS development. We also identify unmet technology requirements that pose barriers to adoption of Deaf-centric TTS technology.

2026-09-09T14:04:39Z Accepted for publication at ACM ASSETS 2026 Shela Atemnkeng Patrick Boudreault Paige DeVries Lloyd May Christian Vogler http://arxiv.org/abs/2609.09719v1 2026-09-09T05:03:12Z

Text-aligned speech tokenization methods have emerged to better align speech tokens with LLM token spaces, enabling more effective utilization of pretrained LLMs. However, they rely on offline automatic speech recognition (ASR), leading to two key limitations: (i) the need for complete utterances before tokenization, precluding real-time streaming, and (ii) vocabulary mismatch between ASR and LLMs, which reduces acoustic granularity from the subword to the word level. We introduce StreamAlign, a text-aligned speech tokenization framework that enables streaming tokenization for real-time speech-text joint modeling. StreamAlign performs online speech-text alignment by combining character-level RNN-Transducer alignment with word-level ASR guidance, mitigating ASR-LLM vocabulary mismatch while preserving recognition accuracy. A proactive word boundary classifier anticipates word completion at chunk boundaries, reducing tokenization latency from 560 ms to 270 ms. On LibriSpeech, StreamAlign achieves the lowest WER and highest UTMOS among evaluated tokenizers. Furthermore, StreamAlign-SLM, a spoken language model trained on StreamAlign units, outperforms other end-to-end spoken language models in speech continuation while achieving the strongest overall consistency on SALMon and spoken StoryCloze.

2026-09-09T05:03:12Z Findings of EMNLP 2026. Project page: https://ishlove77.github.io/StreamAlign/ Kang-wook Kim Jinyoung Park Jinsoo Kim Sehun Lee Sang Hoon Woo Gunhee Kim http://arxiv.org/abs/2609.09677v1 2026-09-09T03:46:14Z

Incremental-text streaming text-to-speech (TTS) needs online text progress tracking for synchronized highlighting, interruption handling, and dialogue-history updates. Input text arrives before it is spoken, so text arrival alone cannot indicate speech progress. Existing waveform-based alignment requires complete audio or adds acoustic processing during streaming. We propose X2-NativeCursor, a lightweight observer that tracks progress from native speech tokens before waveform decoding without changing the TTS generator. Its normalization plan links spoken labels to their original-text spans. Text and native-token encoders feed a local matcher that estimates the current label position. A separate output rule converts revisable position estimates into a cursor that never moves backward. Mean absolute error against an automatic reference is 0.151 Chinese characters with 80-ms lookahead, versus 1.253 characters with 320-ms lookahead for an online waveform baseline. Alignment real-time factor also decreases from 0.3598 to 0.0180 relative to this baseline. Lower tracking error is retained under a second automatic alignment reference. We evaluate X2-NativeCursor on Qwen3-TTS and validate its adaptation to CosyVoice2 by training a separate observer for each backbone. Code is publicly available at https://github.com/X-Square-Robot/X2Streaming-TTS.

2026-09-09T03:46:14Z Zehan Liu Carl Chen Rime Wen Kaiqi Fu Altman Lin Shawn Qin Lights Shi Roy Gan Hao Wang Qian Wang http://arxiv.org/abs/2609.08956v1 2026-09-08T16:11:49Z

Full-duplex voice interaction requires more than utterance-level conversion. It must process streaming speech, manage turn-taking and interruptions, while preserving pretrained linguistic competence and acoustic paralinguistic cues. We ask whether TASTE (Text-Aligned Speech Tokenization and Embedding) provides a viable path toward this goal. We present TASTE2, which transforms utterance-level TASTE into an incremental dialogue stack. A shared text-token vocabulary removes word-level averaging, while modality-aligned dialogue training predicts one continuous audio latent per text token without interleaving heterogeneous token streams. An incremental Speech Detokenizer enables streaming synthesis through CosyVoice2. After speech and dialogue training, TASTE2 (Merge) reaches 56.3% on LLaMA-Questions against a 57.3% Qwen2.5-7B Instruct text-only reference (98.2% accuracy retention), and TASTE2 (Direct) reaches 53.0% (92.4% retention). We build TASTE2 VoiceBot, which processes user speech incrementally, streams synthesized audio, and stops generation on barge-in. On Full-Duplex-Bench v1.0, TASTE2 and TASTE2 VoiceBot handle interruptions well while maintaining high conversational coherence. Natural conversation remains challenging, and deployed mean time to first audio is 2.701 s on two NVIDIA RTX A6000 after TensorRT acceleration. Finally, to our knowledge, we provide the first systematic characterization of explicit paralinguistic control in a TASTE based model. Fast speaking rate serves as a cross-strategy proof of concept after dialogue SFT, while emotion control is strategy dependent and the remaining attributes stay weak. Together, these results establish TASTE based modeling as a practical route toward full-duplex systems while identifying natural conversation robustness, speech generation latency, and feature general paralinguistic control as open challenges. Explore TASTE2 online.

2026-09-08T16:11:49Z Yi-Chang Chen Chun Wei Chen Dien-Ruei Wu Jie Lin Yu-Kuan Fu Yang-Hsien Lin Eddie TC Huang Simon See Hung-yi Lee Da-Shan Shiu http://arxiv.org/abs/2609.08948v1 2026-09-08T16:06:42Z

The rapid advancement of voice synthesis technologies such as text-to-speech and voice conversion poses significant threats to speech-based authentication systems, necessitating robust deepfake detection methods. In this work, we propose a novel E-Branchformer-based architecture that effectively leverages self-supervised speech representations for audio deepfake detection. Our model employs parallel branches to simultaneously capture global contextual dependencies through multi-head self-attention and local temporal patterns through convolutional processing. To enhance discriminative capability, we integrate depthwise convolution and Squeeze-and-Excitation modules that enrich the classification token with refined patch token information after feature merging. Extensive experiments on ASVspoof 2021 LA, DF, and In-the-Wild datasets demonstrate state-of-the-art performance with equal error rates of 0.88%, 1.85%, and 6.30% respectively, substantially outperforming existing methods. Comprehensive ablation studies validate that the dual-branch architecture provides complementary discriminative information, Squeeze-and-Excitation Aggregation significantly improves SSL feature integration, and the combination of DWConv and SE modules is critical for effective class token enhancement. The superior performance on real-world scenarios demonstrates strong generalization capability to diverse acoustic conditions and unseen spoofing attacks.

2026-09-08T16:06:42Z Accepted to SALMA Workshop @ EMNLP 2026 Phuong Tuan Dat Ho Bao Thu Nguyen Tran Trung Pham Viet Hoang Nguyen Thi Thu Trang http://arxiv.org/abs/2609.08703v1 2026-09-08T13:03:46Z

Text-to-speech systems often face a trade-off between natural prosody and efficient inference: higher perceptual quality typically comes at increased computational cost and latency. We present TontaubeV1, a model that preserves natural prosody while enabling streaming from a single consumer GPU. Speech is encoded by the hierarchical DualCodec representation at 12.5 Hz, which separates a semantic stream from successive acoustic refinements. Our design assumes that prosodic structure is largely established when the semantic stream is generated, and allocates capacity accordingly: a Qwen3-1.7B-derived transformer predicts that stream and thereby the utterance duration, while three progressively smaller Qwen3-0.6B-derived transformers each add one acoustic refinement. Text is tokenized per character rather than by subword. Paired text and audio markers at shared positions support long-form generation with bounded context, and overlapping DualCodec reconstructions are mapped into the VibeVoice acoustic latent space and decoded causally, enabling streaming despite DualCodec's noncausal decoder. The model accepts up to one minute of reference audio for voice conditioning and is designed primarily for English and German, with additional multilingual support. The four predictors total 2.9B parameters; on a single RTX 5090 the streaming path reaches approximately 200 ms to first audio. In separate non-streaming measurements, the end-to-end real-time factor (RTF) is 0.08 for one input and the aggregate RTF is 0.02 across eight concurrent inputs. On our LLM-as-a-judge audiobook-reading benchmark, TontaubeV1 matches ElevenLabs Flash v2.5 and outperforms Fish Audio S2 Pro, the April 2026 Gradium API, and Cartesia Sonic 3 on prosody. The model weights are released on Hugging Face under the Tontaube Community Model License 1.0.

2026-09-08T13:03:46Z 13 pages, 2 figures, 2 tables. Both authors contributed equally Fritz Cremer Jonathan Cremer http://arxiv.org/abs/2609.07200v1 2026-09-07T08:24:10Z

Realizing full-duplex spoken dialogue requires large amounts of two-channel, one-speaker-per-channel conversational speech data. Although conversational text-to-speech (TTS) engines have been developed, they are not necessarily robust to two-party simultaneous phenomena such as backchannels, interruptions, and overlaps that occur while the interlocutor is speaking. In this work, aiming at conversational speech synthesis that reproduces human-like overlap, we propose KABURI-TTS. KABURI-TTS takes a per-speaker phoneme raster as input and renders the speech of the two speakers on separate channels, conditioned on the per-frame phonemes and the voice activity derived from them. Because the phoneme raster is supplied by a separate module, the proposed method enables controllable generation of one-speaker-per-channel, two-party spoken dialogue. A user evaluation shows that, compared with strong baselines, the proposed method attains higher naturalness at both the utterance and the interaction level. Furthermore, an analysis of voice activity confirms that the proposed method produces more overlap and more frequent turn-taking.

2026-09-07T08:24:10Z Accepted at APSIPA ASC 2026 Ryuichiro Higashinaka Shinnosuke Takamichi Tetsuji Ogawa