Streaming audio-visual target speaker extraction (AV-TSE) uses a target speaker's face to extract the speaker's speech from an input audio mixture in real time. However, existing systems fix the target speaker for an entire clip, while conversation demands frequent and immediate target switches. To address this problem, we propose Tap2Listen, a streaming AV-TSE model designed for instant response to target switches. The model separates computation by modality: the target-agnostic audio encoder processes only the mixture, while the target-specific visual cues of the selected face condition only the decoder. A switch requires only a 0.12ms re-projection and takes effect immediately. We also design component-wise losses: a prediction loss for the encoder and a supervised extraction loss for the decoder. For evaluation, we measure how quickly output quality recovers after a switch. On the standard LRS3-2mix test set, Tap2Listen improves SI-SNRi from 8.85 to 9.68 dB at a comparable parameter count. After a switch to a known face, the model produces full-quality output from the first frame, whereas the baseline needs about 1 second to recover. Finally, we show that the proposed streaming AV-TSE model runs in real time on a single CPU core.
https://arxiv.org/abs/2610.36643
Reinforcement learning has made Audio Large Language Models (Audio LLMs) better reasoners, but the reward that drives it still comes from outside the model: a final-answer check, a hand-crafted process score, or an external LLM judge that stays fixed while the policy improves. We show that an Audio LLM can supply this reward itself, and can do so recursively: each improved model becomes the critic that trains its successor, so the training signal improves along with the model. We introduce AudioRSI (Recursively Self-Improving Audio Reasoners), a reinforcement learning framework in which a single model plays two roles: as an actor it generates reasoning paths, and as a critic it compares them pairwise, turning its own judgments into a process-level reward. The critic is periodically replaced by the current policy, which closes the recursive loop. Because the critic hears the same recording as the actor, it can reject reasoning that cites acoustic evidence the audio does not contain, rather than reward reasoning that merely sounds plausible. The recursion is what sustains progress: under a frozen external judge, out-of-distribution accuracy plateaus during training, while a critic refreshed from the improving policy keeps it rising. Across four out-of-distribution benchmarks, all disjoint from the AVQA training distribution, AudioRSI outperforms strong baselines including Audio Flamingo 3 and Gemini 2.5 Pro on MMAU test-mini, ranks first on MMSU, achieves a clear lead on the challenging Multi-Audio split of MMAU-Pro, and leads the 7B open-source tier on MMAR among methods without external reward models. With no reward specifying how to reason, AudioRSI also develops structured reasoning behaviors (logical reasoning, perception-aware analysis, and reduced hallucination), corroborated by an external LLM-as-Judge evaluation and a keyword-based behavior analysis.
https://arxiv.org/abs/2610.36578
Real-world audio applications require precise, queryable representations grounded in physical time. Yet modern Large Audio-Language Models (LALMs) increasingly favor conversational generation, producing free-form descriptions that conflate acoustic perception with higher-level reasoning and lack precise temporal grounding. We introduce TAC2, a foundation model that casts timestamped audio captioning as composable functions to construct granular, speaker-attributed acoustic timelines directly from in-the-wild audio. To avoid prohibitively expensive manual annotation, we programmatically fuse complementary predictions from specialized weak teachers through time-registered IoU matching. We model the resulting multi-granularity schema of timestamped events, speaker turns, and fine-grained speech attributes using an Instruction-Input-Output (IIO) completion format, in which each task is a function of the audio, an instruction, and an optional input. A single call decodes the full description autonomously, while composing calls lets earlier outputs or external priors, such as a reference transcript, condition later ones. We further extend TAC2 with expressive speech attributes to obtain TAC2-S, using parameter-efficient specialists to elicit latent paralinguistic capabilities from the pre-trained backbone and distill them into a unified model. We validate the resulting representation in two complementary settings. Paired with frozen text-only LLMs, it yields strong spoken audio understanding across diverse reasoning backends, demonstrating the utility of decoupling fine-grained acoustic perception from downstream reasoning. Used as paired supervision for text-to-audio generation, it supports semantically and temporally grounded reconstruction, together with explicit zero-shot control over event timing, speaker attributes, and word-level emphasis.
https://arxiv.org/abs/2610.36568
Accurate prediction of human Mean Opinion Scores (MOS) is crucial for evaluating synthetic speech quality in text-to-speech (TTS) and voice conversion (VC). Most MOS predictors output only point estimates and ignore uncertainty, limiting reliability in model selection and deployment. We propose ConformalMOS, a framework that augments MOS prediction with conformal prediction to produce statistically valid prediction intervals with guaranteed coverage under exchangeability, alongside point estimates. During training, MOS labels are modeled in an ordinal-aware manner by converting one-hot targets into Gaussian-smoothed distributions, preserving score ordering. This improves uncertainty calibration, particularly when softmax confidence is overconfident on out-of-distribution speech. On BVCC datasets, ConformalMOS achieves competitive accuracy (MSE = 0.08) while delivering empirically validated coverage, enhancing reliability for real-world TTS and VC deployment.
https://arxiv.org/abs/2610.36126
Clinical audio models increasingly align recordings with participant metadata or metadata-derived text. This supervision is useful but ambiguous: the paired target can contain symptoms, demographics, missingness, and cohort structure, not just disease information. We study this ambiguity by making the audio-metadata pairing itself the intervention. Our Pairing-Controlled Transfer Audit trains matched contrastive trajectories under correct pairing, global shuffling, disease-label-preserving shuffling, and label-and-recorded-sex-preserving shuffling. Each endpoint is then frozen and tested with three readouts: profile correspondence, attribute decodability, and matched disease prediction. On UKCOVID, correct improves retrieval and sex decodability over label-only shuffling, showing that the pairing intervention worked. The sex-probe contrast is largely absorbed by the sex-preserving control, while smaller retrieval differences remain. Disease-transfer contrasts, however, remain uncertain. After UKCOVID discovery fixed the pipeline, CODA TB, Cambridge COVID-19 Sounds, and Coswara were rerun as a locked external panel with the same arms, backbones, matching, readout, calibration, primary contrast, and primary metric. The external panel preserves the descriptive separation without proving a universal null. Successful profile correspondence and improved disease transfer are therefore different empirical claims.
https://arxiv.org/abs/2610.35747
Wide-area 6DoF rendering from sparse Ambisonic recordings is commonly addressed by interpolating parametric spatial descriptors rather than the waveforms themselves. This preserves the fidelity of a nearby measured signal and avoids phase artifacts caused by mixing distant arrays. However, existing DirAC-based approaches rely on geometric interpolation weights that do not depend on the acoustic scene observed by the microphones. We propose a lightweight Graph Neural Network (GNN) that refines only this interpolation stage. The known Ambisonic arrays and the query position are represented as a graph whose nodes contain local DirAC features and whose edges encode relative distance and directional relations. The network predicts positive query-conditioned weights over the measured arrays, while the target energy, diffuseness, and direction of arrival are reconstructed by an analytical DirAC layer. Experiments on unseen rooms, unseen speech signals, and unseen array configurations show that the proposed weighting improves energy reconstruction and radiation-pattern similarity with respect to baseline interpolation.
https://arxiv.org/abs/2610.35741
Passive acoustic monitoring in tropical waters such as those in Southeast Asia is hampered by the persistent, high-level impulsive noise of snapping shrimp. This is a particularly daunting challenge for detecting broadband transient vocalizations of dolphins - clicks, buzzes and burst-pulses. These vocalizations, unlike frequency-modulated whistles, share the broadband, impulsive character of the shrimp snaps and cannot be easily separated from them by frequency-localized or energy-based means. We develop TRIDENT, a lightweight waveform-based detector for these vocalizations, and apply it to 54.2 recorder-months of data from ten island sites around Singapore. We show that the training-data pipeline and the architecture are both important: controlled signal-to-noise-ratio based mixing to generate training data, appropriate normalization and hard negative mining, together steer the detector toward learning transient temporal structure rather than energy shortcuts. A single joint detector handles all three signal types better than specialized models for each, and outperforms established baselines. Applied to long-term recordings, it reveals tightly clustered, predominantly daytime activity, motivating days-with-detections as a robust indicator of cetacean presence.
https://arxiv.org/abs/2610.35641
Environmental noise identification is essential for noise control, yet conventional single-stage classifiers suffer from class confusion under imbalanced multiclass conditions. This paper proposes a lightweight hierarchical model integrating a Frequency-Time Adaptive Network (FTANet) with a two-stage classification mechanism. FTANet employs parallel asymmetric one-dimensional convolutions to capture temporal dynamics and frequency-band distributions, while adaptive fusion attention (AFA) enhances discriminative acoustic features. In the first stage, FTANet classifies sounds into five coarse-grained categories covering 25 fine-grained classes. In the second stage, samples are routed to the corresponding category-specific classifiers for fine-grained recognition. On a self-constructed dataset, FTANet achieves 95.45% coarse-grained accuracy with only 0.2004M parameters. On UrbanSound8K, it achieves 79.24% mean accuracy and 80.37% mean F1-score in ten-fold evaluation, outperforming compared convolutional models. Field tests on the Tianjin Online Acoustic Environment Monitoring Platform yield 87.94% coarse-grained accuracy and 97.21% conditional fine-grained accuracy for natural sounds correctly routed to C2. The complete hierarchical model achieves 87.18% overall accuracy and 87.40% weighted-average F1-score, demonstrating its feasibility for urban environmental noise monitoring.
https://arxiv.org/abs/2610.34931
Pretrained image-generation models have been successfully repurposed for various downstream visual tasks. Whether their learned representations and generative priors can transfer beyond vision to other modalities remains largely unresolved, particularly for speech synthesis, which requires linguistic fidelity and preservation of speaker identity from a short reference. We present ImageTTS, which repurposes FLUX.2 for speech synthesis by encoding pseudo-RGB log-mel spectrograms with its frozen image VAE and adapting only its pretrained Diffusion Transformer (DiT) using conditional latent flow matching. This formulation disentangles two forms of vision-to-speech transfer: representation transfer through the image VAE and generative-prior transfer through DiT initialization. The frozen image VAE largely preserves intelligibility and speaker identity with 0.99 STOI and 0.95 speaker similarity, while image-pretrained DiT initialization accelerates convergence and improves linguistic accuracy, reducing WER from 4.2\% to 2.7\% at 100k matched updates relative to random initialization. The same framework supports both zero-shot and instruction-guided TTS and, when scaled to 100K hours of speech data, achieves 1.12\% WER on Seed-TTS test-en and a Style-ACC of 0.66 on CapSpeech, outperforming speech-native baselines trained on comparable data across both tasks. These results suggest that pretrained image models can transfer beyond architectural reuse, supplying useful latent representations and generative priors for intelligible and controllable speech synthesis.
https://arxiv.org/abs/2610.34807
This paper presents S2SND-v2, a unified online and offline speaker diarization framework built on S2SND that improves acoustic feature extraction and the use of the resulting representations. For acoustic feature extraction, we introduce a speaker-adapted w2v-BERT 2.0 front end and compare full and pruned construction routes. To better use these features, overlap-aware activity-query masking improves speaker embedding extraction under overlap, while late-speaker training aligns unknown-speaker training with online inference conditions. Combining these improvements, the pruned system achieves competitive average diarization error rates of 14.52% online and 12.03% offline across eight test sets.
https://arxiv.org/abs/2610.34707
Pretrained audio spectrogram transformers are widely used as backbones of machine listening systems for sound event classification. However, they are typically trained using spectrograms with a fixed Short-Time Fourier Transform (STFT) configuration, implicitly binding the model to the time-frequency resolution seen during pretraining. In practice, adapting these systems to new domains or computational constraints often requires different STFT frontend parameters. In this work, we conduct a controlled study of the sensitivity of the foundational Audio Spectrogram Transformer to variations in STFT window and hop size, showing that audio classification performance degrades substantially as frontend parameters deviate from pretraining settings. As a mitigation strategy, we introduce a lightweight conditioning mechanism using FiLM that enables a pretrained model to adapt to variable input configurations using limited additional training data. Our method generalizes well to unseen STFT configurations, improving model robustness without sacrificing performance in the original setting.
https://arxiv.org/abs/2610.34533
Kernel distances such as the Kernel Inception Distance (KID) and the Kernel Audio Distance (KAD) compare generated and reference samples in an embedding space, and model comparisons rest on their values. Each ships with one fixed kernel, and a reported value does not say which changes that kernel detects at the sample size used. We run each default as a permutation two-sample test on real embeddings of symbolic music, audio and images, comparing a sample with an independent sample changed by a mean shift, an isotropic contraction, or a rotation that keeps the mean and the covariance trace and is matched to the shift in transport distance. At 780 samples per side, both defaults rarely detect the rotations in five evaluation spaces, while they detect the shifts in every run. The median-bandwidth Gaussian kernel of KAD detects the contraction and regains power against rotations on image embeddings with an order of magnitude more samples. The cubic kernel of KID contains second- and third-moment information but misses the contraction in most runs, and our experiments do not establish a high-power rotation regime for it over the tested sample sizes. Narrower Gaussian bandwidths restore sensitivity in music and audio, an $\ell_1$ Laplace kernel does so for images, and aggregated tests combine both at nominal level. We recommend that a reported kernel distance state its sample budget and the changes its configuration detects at that budget.
https://arxiv.org/abs/2610.34477
While recent advances in text-to-speech (TTS) synthesis have substantially improved speech quality and modeling flexibility, the development of lightweight and computationally efficient systems remains comparatively underexplored. We present LeanTTS, a bilingual complete synthesis stack that supports two complementary regimes: reference-prompt zero-shot cloning for unseen speakers and token-conditioned customization for recurring voices. LeanTTS combines a compact Mandarin polyphone front end, a structurally compressed flow-matching acoustic model, and a lightweight Vocos decoder with four-step MeanFlow generation. The resulting system contains 18.9M learned parameters, 86.12% fewer than the 136.2M-parameter ZipVoice-Vocos pipeline. On Seed-TTS-Eval, its zero-shot mode obtains 2.02 CER/62.3 SIM on Mandarin and 2.32 WER/48.3 SIM on English, exposing the quality-size trade-off under aggressive compression. With token-conditioned customized voices, the same system reaches 1.57 CER/77.2 SIM on Mandarin and 1.93 WER/76.6 SIM on English, with UTMOS scores of 3.03 and 4.33, respectively. It further achieves mean/P90 CPU RTFs of 0.049/0.061, establishing a sub-20M operating point for bilingual personalization and real-time CPU deployment. Audio demos are available at this https URL.
https://arxiv.org/abs/2610.34408
Spatial audio is central to immersive 360$^\circ$ media, yet existing audio-visual spatialization renders the scene as a whole, leaving no control over an individual source. We introduce Audio-Visual Query-conditioned Target Sound Spatialization (AV-QTSS): given a monaural multi-source mixture, a 360$^\circ$ video and a class query from the 13-class DCASE sound-event ontology, it generates the first-order Ambisonics (FOA) signal of the queried source alone. The omnidirectional mixture contains no explicit spatial cues, so direction has to be inferred from the video, where several objects may sound at once. We propose a query-conditioned dense binding localizer that predicts a spherical probability map over the whole scene, coupled to the renderer so both are optimized jointly. Experiments with one to four concurrent sources show improved localization and spatial accuracy relative to established methods and ablated variants, and the localizer transfers to real 360$^\circ$ recordings after adaptation.
https://arxiv.org/abs/2610.34401
Symbolic music representations encode musical content at varying levels of completeness, from performance-oriented formats such as MIDI to fully notated scores such as MusicXML. Yet the impact of this representational richness on deep learning models remains poorly understood. While tokenisation has been extensively studied for MIDI, score-based formats have received considerably less attention, and MusicXML in particular remains largely unexplored despite being the de-facto standard for score interchange. In this paper, we introduce the first general-purpose tokenisation framework for MusicXML, and conduct a systematic empirical study to determine whether notational richness confers a measurable advantage for deep learning on symbolic music. We train 26 autoregressive and masked language models, comparing MusicXML tokenisation against MIDI and ABC-based representations, probing the effect of different tokenisation strategies within MusicXML, and testing whether music-grammar-aware tokenisation outperforms treating scores as plain text. We evaluate music understanding tasks using linear probing across composer classification, difficulty estimation, and accompaniment suggestion, and generative tasks using objective statistical metrics and a human pairwise preference study. Our results show that MusicXML-tokenised models outperform MIDI and text-based counterparts on understanding and generative tasks.
https://arxiv.org/abs/2610.34247
Learning audio representations that effectively bridge holistic global soundscapes with fine-grained acoustic moments is a long-standing goal in self-supervised learning. Recording-level contrastive objectives encourage agreement between pooled audio representations, but do not explicitly supervise local time-frequency representations. We introduce AudioFM, a self-supervised framework that combines recording-level contrast with patch-wise masked latent discrimination. Using visible acoustic context, a student predicts representations of masked patches and identifies their corresponding full-view teacher targets among candidates from other recordings. We further incorporate weaker supervision from eligible temporal neighbors to encourage local temporal consistency. Extensive evaluations show that AudioFM achieves state-of-the-art performance across diverse audio domains in linear probing, fine-tuning, and zero-shot transfer, demonstrating superior event detection, robust scene-level understanding, and strong domain generalizability.
https://arxiv.org/abs/2610.34183
Existing anti-spoofing systems report either an utterance-level score or a time-based spoof interval: the former cannot show which words are spoofed, while the latter does not directly match the transcript and must be post-aligned with words, so alignment errors can lead to wrong word-level labels. Standard ASR model outputs only text; spoof information in the audio is lost after recognition and is not passed with the transcript. Yet changing only a few words can change the meaning of a sentence, leaving downstream systems to use incorrect text without knowing which words are spoofed. We therefore treat partial anti-spoofing as tagged ASR: the model transcribes speech and marks spoofed words at the same time, directly linking spoof labels to output words. We adopt the audio LLM and use ASR-Protected Subspace Gated Fusion (APSGF) to fuse spoof cues with ASR features. Spoof cues are added only outside an ASR-protected subspace, so they do not disturb recognition features. After cross-entropy training, we fine-tune the model with Reference-Anchored Policy Optimization (RAPO). Since the SFT policy is already concentrated, multiple samples from the same input tend to receive similar rewards, resulting in weak group-wise advantages. RAPO instead generates a single sample with low sampling randomness and anchors its advantage to the reward of a reference model. This allows the model to reinforce responses that outperform the reference while suppressing those that underperform it. Experiments on five datasets demonstrate the effectiveness of our approach in jointly detecting spoofed words and preserving ASR accuracy. The code is publicly available.
https://arxiv.org/abs/2610.34156
Speech Emotion Recognition (SER) is a fundamental task in affective computing that enables applications such as mental health assessment, human-computer interaction, and automated customer support systems. Accurate emotion classification from speech remains challenging due to variations in speaker style, acoustic conditions, and the high computational cost of modern deep learning approaches. Many existing systems depend on large transformer-based architectures that require substantial computational resources, making them less suitable for real-time or resource-constrained environments. This research proposes EmoNet, a lightweight deep learning framework for speech emotion classification across seven emotion categories. The model is built on a four-block depthwise separable CNN encoder with squeeze-and-excitation attention and residual connections, combined with a Bidirectional LSTM for temporal feature learning and a multi-head attention pooling mechanism for improved representation. A comprehensive 134-channel acoustic feature set alongside data augmentation techniques including audio-level perturbations and SpecAugment is employed to improve robustness under diverse acoustic conditions. Evaluated on a combined dataset of TESS, RAVDESS, and SAVEE, EmoNet achieves a test accuracy of 96.75% with only 4.11 million parameters. The proposed model outperforms its reproduced baseline and remains competitive with significantly larger state-of-the-art architectures.
https://arxiv.org/abs/2610.33977
Sperm-whale codas are usually described by the number and timing of their clicks. Earlier acoustic studies showed that individual clicks and longer coda sequences contain additional organization. We ask how that organization compares with music, whose smaller sounds also form larger units and ordered sequences. We compare codas with performed violin music, drum grooves, and sung melodies. Their sequences resemble music in some measures of predictability, yet codas show a different relationship between the smaller sounds within a unit and the order of units across a sequence. This combination was not observed in the musical references tested. Coda timing also varies coherently and carries information about individual callers. The recordings show another striking property: broad, reproducible variation along acoustic dimensions associated with human ratings of emotion, with a range comparable to human audio in most eligible tests. These findings motivate an acoustic blueprint that describes natural codas beyond their click timing. Its importance is apparent in synthetic audio: WhAM produces recognizable coda-like recordings, yet the tested clips remain distinguishable from natural codas in five of six acoustic comparisons after strict matching. The blueprint provides a fuller way to study natural codas and to evaluate how faithfully synthetic recordings represent them.
https://arxiv.org/abs/2610.33974
Automatic DJ systems mostly target electronic dance music, whose regular structure makes rule-based mixing relatively easier. Vocal-rich popular music is harder: which track to play next, where to cue it, and which transition to use are tied together, so fixed rules handle them poorly. We present DJustify, a pop AI-DJ that puts an LLM in a reason--propose--verify--repair loop and lets it choose the next track, cue points, and mixing technique together. The LLM samples several transition plans; an external acoustic checker then scores the rendered audio and triggers repair when a check fails. Sampling keeps several candidate transitions per track pair instead of one fixed mix. In blind listening tests our transitions beat Mosaikbox, a reimplemented crossfade, and a cue baseline that reuses the crossfade's cues; that last comparison shows most of the margin comes from cue placement. The same planner also picks and orders a set along an emotion arc: this improves story coherence over a no-story ablation, and within a set story and transition ratings show no measurable gap from a human DJ. The checker mainly adds reliability, not higher average quality: it repairs rare but audible errors such as cut vocals, while the planner stays non-deterministic.
https://arxiv.org/abs/2610.33434