S Watanabe

Shikhar Bharadwaj, Kwanghee Choi, Stephen McIntosh et al. · cmu

Phone segmentation and recognition are inherently related tasks, yet modern approaches typically model them separately. We argue that phonetic structure is already latent in the representations of self-supervised speech models (S3Ms), and one only needs to steer them to solve both tasks. We leverage S3M-based Phonological Activation Mapping (SPAM), which maps each S3M representation frame to a vector of phonological feature activations, such as voicing and nasality. On top of SPAM, we introduce two simple but effective lightweight, gradient-descent-free prediction heads: a recognition head and a segmentation head. Our method requires less than a minute of phonetic transcriptions, and generalizes to unseen phones during training. Across a diverse range of datasets, our approach attains strong segmentation and recognition performance.

Haoran Wang, Jinchuan Tian, Siddhant Arora et al.

While Large Multimodal Models excel in comprehension, high-throughput inference engines lack native support for multimodal generation. This is severe in Speech Language Models, where generating multi-layered audio tokens via decoupled AR+NAR or synchronous Multi-Token Prediction (MTP) with delay-pattern interleaving conflicts with standard single-stream loops. We present a vLLM-based inference pipeline for unified speech understanding and generation. We extend autoregressive decoding to natively execute delay-pattern de-interleaving and coordinated multi-stream sampling, integrating an on-GPU acoustic decoder for end-to-end waveform synthesis. Crucially, we overcome the shared intuition that Classifier-Free Guidance (CFG) halves throughput. By co-scheduling paired conditional and unconditional requests within a continuous batch, our CFG implementation sustains 80% of non-CFG throughput, absorbing dual-request and logit merging overheads. We open-source our framework.

Xun Gong, Jinchuan Tian, Haoran Wang et al.

Current text-guided audio editing methods rely on paired training data, predefined operation templates, and separate processing pipelines across speech, music, and sound. We present Bagpiper-Edit to enable open-ended audio editing via free-form natural language instructions. We reformulate audio editing as a rich-caption rewriting task by treating a rich caption as the semantic representation of an audio clip. The user request is translated into an edited caption, which then guides Bagpiper-Edit to generate the target edited audio with the original audio as contextual acoustic anchor. This unlocks the potential of free-form editing, and circumvents the need for paired audio-editing training data, enabling powerful zero-shot editing capabilities. Evaluations across speech, audio, and free-form editing show Bagpiper-Edit maintains good consistency to the original audio and achieves similar performance to other expert models in most cases. Demo: https://bagpiper-edit.github.io, Codes: https://github.com/espnet/espnet/pull/6417 & https://github.com/HsunGong/espnet

Nathan Pruyne, Benjamin Stoler, William Chen et al.

We present MulTTiPop, a dataset of pop music segments and their associated multitrack MIDI recordings for the evaluation of automatic music transcription models. MulTTiPop contains 572 segments of popular music totaling 3.5 hours of audio, and contains songs from diverse genres and decades from the 1930s to 2000s. To collect this dataset, we perform metadata-based matching on song segments from the Lakh MIDI and TheoryTab datasets, manually identify an anchor beat between the audio and MIDI, then use beat tracking on the audio and warp the MIDI to match its tempo and timing. We evaluate state-of-the-art automatic music transcription models on MulTTiPop and find substantial room for improvement, with the best model achieving 38% Onset F1. More details and sound examples of MulTTiPop are available at https://gclef-cmu.org/multtipop.

Yohsuke Tsujino, Shun Watanabe

The modulo-sum problem, proposed by Körner and Marton (KM), is a representative problem in the field of distributed source coding. In the modulo-sum problem, two correlated sources are encoded separately, and the decoder decodes the modulo-sum of the sources. It is clear that the Slepian-Wolf (SW) coding rate region is achievable for the modulo-sum problem. Körner and Marton proved that the SW coding rate region can be improved by structured coding based on linear codes. Ahlswede and Han (AH) proposed AH coding, which combines structured coding and random coding, and expressed its rate using auxiliary random variables. However, it was conjectured that the minimum sum rate of AH coding cannot be smaller than the minimum of the sum rates achievable by KM coding or SW coding. Subsequently, Kakishima and Watanabe considered a multi-letter extension of AH coding, and designed the auxiliary random variables by taking the XOR of adjacent bits of the source sequences. Through numerical computation, they demonstrated the existence of source parameters for which multi-letter AH coding improves upon SW coding. However, this confirmation remained numerical, and the conditions under which multi-letter extended AH coding improves upon SW coding have not been analytically characterized. In this study, we analytically characterize the conditions under which multi-letter extended AH coding improves upon SW coding. Our condition is tight in the sense that it coincides with the complement of the known SW optimal sufficient condition. To obtain the conditions, we apply the method of types to reduce the evaluation of the multi-letter expression to a comparison of single-letter divergences, which might be of independent interest.

Abinay Reddy Naini, Jaeyeon Kim, Chao-Han Huck Yang et al.

Large audio-language models (LALMs) can reason about audio, yet it remains unclear whether they can perform comparative judgments between two speech signals along emotional, environmental, linguistic, prosodic, and interpersonal dimensions. We study this question in the context of speech emotion recognition (SER), where the model determines which utterance exhibits higher arousal, valence, or dominance. We introduce a reasoning-guided ordinal SER framework that conditions an LALM on paired speech inputs. The model is trained using reasoning traces generated from both semantic audio descriptions and acoustic evidence derived from GeMAPS features, enabling interpretable comparative decisions. Beyond direct supervision, we also employ direct preference optimization to encourage stronger separation for emotional differences. Experiments show that the proposed framework improves preference prediction while requiring only 5% of the training data used by conventional ordinal SER systems.

Jinchuan Tian, Haoran Wang, Siddhant Arora et al.

Classical TTS systems typically rely on rigid input formats and predefined metadata slots, limiting their ability to fulfill flexible user requirements. This paper introduces Bagpiper-TTS, a universal speech synthesis system that deals with diverse natural language user requests. Given a natural language prompt, Bagpiper-TTS first reasons over the users' intent to derive a rich caption, i.e., a comprehensive textual blueprint encompassing both transcription and nuanced metadata. Subsequently, this caption guides the synthesis of the target speech. Our model inherently supports a broad spectrum of tasks besides classical TTS applications, including multi-talker, intent-to-speech, role-play synthesis, singing voice synthesis, and more. Experimental results demonstrate that Bagpiper-TTS achieves an 1.7% Word Error Rate (WER) on the Seed-TTS-Eval benchmark and match the performance of dedicated models in both LLM-as-a-judge and human subjective evaluations across multiple applications.

Masao Someki, Alexander Polok, Carlos Carvalho et al.

Recent speech research involves increasingly large datasets, complex models, and diverse experimental workflows. However, existing frameworks require substantial engineering effort to support such experiments. We present ESPnet3, a speech and audio research framework built on a modular system architecture with configuration-driven dataset composition and unified Python-based workflows. ESPnet3 introduces a DataOrganizer abstraction for flexible dataset integration and dataset sharding for memory-efficient large-scale training, while allowing recipe-specific logic through lightweight stage overrides. In OWSM pre-training experiments, ESPnet3 reduces per-epoch training time by \emph{21.1 minutes} compared to ESPnet2 and achieves \emph{>80\% GPU utilization} in multi-node training. Fine-tuning experiments show that new models and datasets can be integrated with around \emph{46 lines of additional code}. ESPnet3 will be publicly released with model checkpoints and training logs.

Keita Goto, Takashi Maekaku, Jin Sakuma et al.

Dual-mode self-supervised speech models are pre-trained to handle streaming and non-streaming conditions simultaneously. However, their attention is computed over different context ranges, which often makes optimization difficult. In previous work, we proposed online registers, additional tokens intended to compensate for missing future context in streaming mode, but the gains remained limited. To address these issues, we introduce two improvements for robust dual-mode pre-training: (1) Online Predictive Coding (OPC), which regularizes the registers through multi-step future prediction, and (2) Dual-mode Layer Normalization, which stabilizes optimization. We fine-tune the proposed dual-mode self-supervised speech models for speech recognition on LibriSpeech and WSJ. Results show that OPC consistently reduces the online-offline performance gap; at 160 ms latency on LibriSpeech, word error rates improve from 3.65% to 3.40% on test-clean and from 10.15% to 9.65% on test-other.

Ryo Fukuda, Takatomo Kano, Siddhant Arora et al.

We investigate turn-taking in multimodal multi-party conversations using large language models (LLMs). We construct an evaluation framework for three tasks: addressee detection, turn-change prediction, and next speaker prediction. We compare supervised models trained for these tasks, text-based LLMs, multimodal LLMs (MM-LLMs), and human subjects. Experiments on the AMI corpus showed that LLMs outperformed supervised models and humans in next speaker prediction, despite not being trained on the target domain and without access to audio or visual information. An MM-LLM performed better than text-based LLMs on addressee detection and turn-change prediction but remained below human performance, indicating difficulty leveraging raw audio-visual signals. Ablation analyses revealed that conversational context was critical, particularly for next speaker prediction. We observed that human and LLM prediction patterns were similar, and intervals with frequent turn changes were difficult for both.

Sathvik Udupa, Shinji Watanabe, Petr Schwarz et al.

While low-latency interaction is critical for spoken dialogue, cascaded architectures are often bottlenecked by reactive turn-completion detection. We propose Endpoint Anticipation, shifting from reactive detection to proactive forecasting of end-of-turn signals. Our speech-based model anticipates endpoints upto 2.56 seconds in advance, enabling speculative execution of LLM and TTS pipelines on partial context. We introduce metrics to quantify the trade-off between realized latency reduction and computational redundancy. Evaluation across conversational and task-oriented datasets shows our model consistently outperforms competitive VAP-based baselines. Integration with the Unmute framework demonstrates a 505 ms average latency reduction with a 28.4% increase in speculative computation, effectively masking sequential bottlenecks to enable complex reasoning in real-time speech-to-speech interaction.