SonoWorld4D: From One Video to a 4D Audio-Visual Scene via Diffusion Synchronization
Derong Jin, Xiyi Chen, Ming C. Lin, Ruohan Gao
AI 速览：提出从单段单目视频生成可探索、随时间变化的4D音视频场景的框架，并引入Video-to-4DAVScene新任务。
Recent progress in visual scene generation has extended beyond static 3D scenes toward the more challenging problem of generating dynamic 4D scenes from minimal input. However, sound, despite being intrinsically coupled with dynamic objects and events, remains largely unmodeled in prior work. We introduce Video-to-4DAVScene, a new task that turns a video into an explorable, time-varying 4D audio-visual scene, and present the SonoWorld4D framework for this task. Given a monocular video, our method generates a 360 static background, reconstructs explicit motion trajectories of dynamic, sound-emitting foreground objects, and synthesizes a physically-grounded sound field consistent across multiple sources through dual-space audio diffusion synchronization. The resulting 4D scene supports real-time novel-view visual and spatial audio rendering. Evaluation on real-world dynamic scenes and a user study demonstrates SonoWorld4D generates 4D audio-visual scenes with perceptually realistic, physically grounded spatial audio consistent with scene geometry and visual content.
Discovering Interpretable Acoustic Features in Autoregressive Voice-Cloning TTS Models
Anuj Diwan, David Harwath, Eunsol Choi
AI 速览：该论文提出一个基于特征的框架，用于解释自回归语音克隆TTS模型内部表征的声学属性，并评估稀疏自编码器、PCA和随机方向在探测与操控上的效果。
Voice-cloning TTS models synthesize an unseen speaker's voice from brief speech prompts, but it remains unclear which acoustic properties they represent and use. We introduce a feature-based interpretability framework for autoregressive voice-cloning TTS models, analyzing features from sparse autoencoders (SAEs), PCA, and random directions via sparse probing, steering, and quantitative auto-interpretability. We focus on Qwen3-TTS and extend probing and steering to $5$ additional models to test generalization across architectures. These models encode concepts like pitch, energy, noise, and accent, but sparse prediction and steering are effective only for some concepts. Strong probing does not always translate to steering or vice-versa. SAEs are useful for sparse probing of human-labelled concepts and novel-concept discovery, while PCA is competitive for dense probing and steering. Auto-interpretability finds $684$ automatically validated SAE features in Qwen3-TTS covering speaker clusters, speaking style, and recording conditions beyond known concepts. Overall, we perform a comprehensive interpretability study of autoregressive voice-cloning TTS models.
Bridging Neural Representations and Symbolic Semantics in Audio Classification
Avishek Das, Pascal Hitzler
AI 速览：该论文提出一个模型无关的神经符号框架，用符号概念归纳分析音频分类器最后隐藏层的32个神经元，探索其能否对应人类可理解的声学概念。
Deep audio classifiers can achieve strong predictive performance, but their hidden representations remain difficult to interpret. Existing audio explainability methods often focus on input-level attribution, such as waveform segments, spectrogram regions, or acoustic features, and provide limited insight into what hidden neurons encode. Interpreting hidden neurons at the concept level can provide a more meaningful view of what an audio model learns internally. This paper investigates whether hidden neurons in audio classifiers can be associated with human-understandable symbolic acoustic concepts. A model-agnostic neurosymbolic framework is presented for analyzing the final hidden layer of audio classifiers, containing 32 neurons, through symbolic concept induction. The framework is evaluated on two audio classification settings: music genre classification using GTZAN dataset and musical instrument classification using IRMAS dataset. Results show that several hidden neurons can be associated with meaningful induced concepts, and most confirmed neuron-concept pairs are statistically supported using the Mann-Whitney U test. These findings suggest that symbolic concept induction is a promising direction for explaining latent audio representations beyond input-level attribution.
What Makes a Good Audio Prompt for Zero-Shot TTS? Learning Prompt Selection from Synthesis Outcomes
Arnon Turetzky, Yossi Adi
AI 速览：作者分析大量零样本 TTS 合成输出，发现提示片段选择显著影响合成质量，并提出一个仅凭音频即可对候选提示排序的小型网络。
Zero-shot text-to-speech (TTS) models clone a target voice from a short reference clip, commonly referred to as the prompt. In practice, the prompt has to be cut from a longer recording of the speaker, so which part to use must be chosen. To understand what makes a good prompt, we analyze hundreds of thousands of outputs synthesized by four TTS models from candidate prompts of YouTube recordings. The choice of prompt substantially changes intelligibility, naturalness and speaker similarity. Standard predicted-quality scores capture mainly the naturalness part but miss most of the intelligibility gain. We find that the models largely agree on which prompt yields the highest measured speaker similarity, but much less on which prompt yields intelligible speech. We propose a small network that learns from the outputs a TTS model produced to rank a recording's candidate prompts from their audio alone. Without synthesizing anything at selection time, it picks prompts that lower the word error rate on held-out recordings by a third (Qwen3-TTS) to two fifths (F5-TTS) relative to a randomly chosen prompt, more than predicted quality achieves and nearly as much as synthesizing a dozen test sentences from every candidate first.
Symmetric Structure and Motivic Transformation in Melodies, Phrases, and Themes
Alfredo González-Espinoza
AI 速览：作者用移位、倒影、逆行与循环移位生成的群，对852首键盘作品的旋律窗口做轨道与稳定子分析，量化旋律的动机词汇、部署与对称性。
Transposition, inversion, and retrograde are fundamental operations in music theory, from Schoenberg's twelve-tone technique to Forte's set classes and Lewin's transformational theory, yet their use in melodies has rarely been quantified. We read a melodic line through the orbits and stabilizers of its windows of $m = 3-6$ notes under the group generated by these operations and cyclic shift. The window entropy splits into an orbit entropy (motivic vocabulary) and a within-orbit entropy normalized by orbit size (deployment), and the group elements relating windows of one orbit give the motivic transformations. In 852 keyboard pieces spanning four centuries, at least 97% of melodies have lower orbit entropy than surrogates preserving their pitch content, spectrum, or one-step transitions ($m \geq 4$). Periods differ in vocabulary and symmetry types; musical forms differ in deployment. Recurring motifs are transposed 1.6 times more often than in random permutations of their forms, and transposed by a fifth or fourth 2.8 times more often, a preference that weakens from the Baroque to the 20th century and from Bach's fugues to Shostakovich's. Folk phrases and instrumental themes are rarely symmetric (2-10%), but are 1.6-5 times more often symmetric than note rearrangements. The framework extends to any symbolic sequence with a symmetry group.
Thomas A Rushton, Romain Michon, Tanguy Risset, Clemens Wegener, Pierre Lecomte
This paper presents an accessible, distributed approach to higher order ambisonics using a network of low-cost microcontroller-based devices. Leveraging recent advances in embedded computing, networked audio, and precise time synchronization, the system offers an alternative to conventional centralized architectures that typically require expensive multichannel interfaces and high-performance digital signal processing hardware. The proposed architecture combines an audio plugin for soundfield design with a modular network of ambisonics processing units, which, exploiting the parallelisable nature of ambisonics decoding, are each responsible for a subset of loudspeaker outputs. Synchronization is achieved through the IEEE 1588 Precision Time Protocol, complemented by adaptive clock correction to maintain sub-microsecond alignment between devices. A third-order ambisonics implementation demonstrates the viability of the approach, with potential scalability to higher orders and larger arrays. Overall, this work contributes to the development of frugal, distributed solutions for immersive audio, expanding access to ambisonics-based systems for a wide range of applications.
We present a cascade streaming voice privacy system, consisting of ASR, NLP, and TTS modules. By leveraging recent improvements in all three aspects of speech and language processing, we show that such a system is capable of achieving a result comparable to state-of-the-art offline voice privacy systems, while providing controllable capabilities of content protection beyond voice. With an estimated utterance-onset end-to-end latency of 717 ms, the presented system preserves speech content with 11% WER on Librispeech, while detecting PII with a 69% SLU F1 score and achieving up to 47.66% EER against a semi-informed attacker.
Arne Hildenbrand, Philipp Diel, Zhuo Cao, Hanno Scharr, Stefan Pischinger, Steffen Marburg
In engineering, applications of artificial intelligence have so far been largely limited to surrogate models and optimization tasks. Large language models changed this, as they can write software from a task given in words. Placed inside a harness, a system of components that lets it search for information and execute software, such a model becomes an agent that can take over recurring engineering tasks. This work measures how the model and the elements of the harness shape the performance and behavior of the agent in two common acoustic applications. In the first, the agent detects and locates faults in audio recordings from a vehicle interior with near-perfect accuracy. A benchmark on this task shows that a weak model sets a limit that no harness design can overcome, and that design choices take effect only once the model is capable enough to use them. In the second, a case study on the design of a metamaterial unit cell, the agent reliably finds designs that achieve a specified insertion loss over a given frequency range. The results aim to help the reader understand what each element of an agent contributes and how to design an agent for acoustic tasks.
Muhammad Sharif, Zeeshan Abbas, Jiangyan Yi, Chenglin Liu
Automatic Speech Recognition (ASR) technology has witnessed significant advancements in recent years, revolutionizing human-computer interactions. While major languages have benefited from these developments, lesser-resourced languages like Urdu face unique challenges. This paper provides an extensive exploration of the dynamic landscape of ASR research, focusing particularly on the resource-constrained Urdu language, which is widely spoken across South Asian nations. It outlines current research trends, technological advancements, and potential directions for future studies in Urdu ASR, aiming to pave the way for forthcoming researchers interested in this domain. By leveraging contemporary technologies, analyzing existing datasets, and evaluating effective algorithms and tools, the paper seeks to shed light on the unique challenges and opportunities associated with Urdu language processing and its integration into the broader field of speech research.
Word-level emphasis conveys meaning that text alone cannot express and is therefore essential for a practical text-to-speech (TTS) system. In this work, we study how to add emphasis control post hoc to a pretrained TTS model that has never been trained with emphasis annotations. We present EmpDPO, a simple, efficient, and effective two-stage recipe that adds emphasis control to a multilingual zero-shot diffusion TTS model through supervised fine-tuning followed by direct preference optimization (DPO). The training signals for both stages come from an off-the-shelf multilingual emphasis detector, so that no manual emphasis annotation or task-specific reward design is needed. We find that DPO can substantially strengthen emphasis control with only a few thousand preference pairs and training steps. Objective and subjective evaluations against strong public systems show that our model produces perceptible emphasis while preserving naturalness and overall quality.
Abishek Soti, Thura Pyae Sone, Naqib Ibnul, Jim Huang, Sergio Chevtchenko, Henry Zhong, Gregory Cohen, Ying Xu
We propose a Raspberry Pi-based underwater vessel classification system for edge deployment under limited connectivity, bandwidth and power in offshore acoustic surveillance. Using ShipsEar, we compare six classifiers on a Raspberry Pi 5 through a complete audio-to-prediction pipeline. We assess classification performance, processing latency, CPU utilisation, memory consumption and monitored power. Supporting experiments examine acoustic feature selection, statistical pooling and Transformer patch configurations. The two-layer convolutional neural network (CNN) achieved a mean macro-F1 of \textbf{(99.14 \textpm{} 0.32)\%} across five random seeds. The radial basis function support vector machine (RBF-SVM) achieved \textbf{98.83\%}, with a median end-to-end latency of 28.109~ms compared with 120.655~ms for the CNN. This favourable trade-off motivated RBF-SVM deployment on the Raspberry Pi Zero 2 W. It maintained the same classification performance with a median end-to-end latency of \textbf{349.09~ms} and a sustained throughput of \textbf{2.962 classifications/s}, or \textbf{14.8} times the rate required for consecutive five-second audio segments.
Listening to one chosen talker in a crowded room is effortless for people but hard for small devices, whose microphone arrays are too compact for classical beamforming to separate nearby voices. We present IsoNet, a select-to-listen system: the user selects a face in the camera view, and a 4-microphone array with a 9.4 cm aperture returns that person's speech. IsoNet fuses multi-channel complex spectra, GCC-PHAT spatial cues, and a face embedding at the bottleneck of a U-Net mask estimator, trained with auxiliary direction-of-arrival supervision and an SNR curriculum on simulated reverberant VoxCeleb mixtures. On a hard test set spanning -1 to 10 dB SNR, IsoNet reaches 9.31 dB SI-SDR, a 4.85 dB improvement over the mixture, and raises PESQ from 1.40 to 2.13 and STOI from 0.72 to 0.84. The gains are largest where the task is hardest (5.71 dB SI-SDR improvement at -1 to 1 dB SNR), and IsoNet outperforms oracle-steered delay-and-sum and MVDR beamformers by 9.7 and 10.9 dB SI-SDR. Ablations show that visual and spatial cues are complementary, and a 4.9M-parameter audio-spatial variant stays within 0.16 dB of the full model while running 26 times faster than real time.
Phu-Anh-Tai Do, Huu-Hoang Nguyen, Duc-Hoang Nguyen, Thanh-Hai Le
Automatic Music Transcription (AMT) seeks to convert acoustic music recordings into symbolic musical notation, a process that remains remarkably difficult for polyphonic piano due to concurrent polyphony, harmonic partial interference, expressive timing, and severe pitch activation sparsity. In this paper, we formulate polyphonic piano transcription as a frame-level multi-label classification problem across the standard 88 piano pitches (MIDI notes 21 to 108). Raw audio waveforms are processed using the Constant-Q Transform (CQT) to construct a geometrically spaced log-frequency time-frequency representation that aligns with the chromatic musical scale across seven octaves (252 frequency bins per frame). We systematically investigate recurrent sequence models, contrasting unidirectional Long Short-Term Memory (LSTM) with Bidirectional LSTM (BiLSTM) backbones, and propose two architectural adaptations tailored to recurrent temporal representations: a modified additive attention module and a modified multihead attention mechanism equipped with hyperbolic tangent activation and head-wise summation. Furthermore, to mitigate the severe class imbalance between inactive and active notes (approximately 41:1), we evaluate Weighted Binary Cross-Entropy (WBCE) alongside standard Binary Cross-Entropy (BCE). Rigorous empirical evaluations on the MAPS and MAESTRO datasets demonstrate steady, compounding gains from bidirectionality and attention mechanisms. Our proposed Multihead Attention-BiLSTM architecture attains an F1-score of 88.20%, an accuracy of 78.90%, a precision of 0.89, and a recall of 0.86, substantially outperforming the baseline unidirectional LSTM (F1-score of 71.20%) while maintaining a compact parameter budget of 4.43M. Qualitative piano-roll visualizations confirm that our model generates faithful transcriptions even in complex harmonic passages.
Few-shot voice cloning makes it cheap to produce daily Japanese financial commentary with several synthetic speakers, but kanji words with more than one reading, such as the word for "market" (shijo vs. ichiba) or "overall trend" (taisei vs. oze), are still misread. A common quality gate transcribes each synthesized sentence with automatic speech recognition (ASR), converts both the input and the transcript to kana, and regenerates on mismatch. We show that this gate is unreliable in both directions: when the audio contains a misreading, the ASR language model frequently writes the expected kanji back, hiding the error, and when the audio is correct, ordinary recognition errors trigger needless regeneration. We propose a reading-level check that scores the expected reading and its confusable alternatives with a connectionist temporal classification (CTC) hiragana recognizer and flags a word when an alternative explains the audio better. On 540 utterances from three voice-cloned production voices covering 30 financial terms, with misreadings injected under controlled conditions, the transcript-based gate detected 78.9% of misreadings at a 44.4% false-alarm rate, whereas the proposed check detected 91.7% at 5.0% (AUC 0.990). On natural kanji input, it caught a front-end error that the transcript-based gate missed for one voice, and reduced regeneration requests from 38.9% to 6.1% of sentences.
Mercy Muchai, Kevin Chege, Nick Mumero Mwangi, Samuel Chege Maina, Felermino Ali, Elizabeth Ankrah, Ignatius Ezeani
Despite recent advancements in ASR, identifying effective architectures, fine-tuning approaches, and data adaptation strategies for low-resource languages remains challenging. We present a systematic comparison across five low-resource African languages and Swahili as a mid-resource contrast, fine-tuning three models from distinct architectures: Phi-4-multimodal-instruct, Whisper-large-v3-turbo, and MMS-1B-all. We analyze data sufficiency, monolingual versus multilingual adaptation, domain fine-tuning, and language-adapter transfer. Phi-4 achieves the best overall accuracy, MMS shows comparable CER, while Whisper exhibits high CER, consistent with known hallucinations and repetition loops. Performance saturates at different data scales, offering practical data-efficiency targets, while typological complexity consistently shapes error rates. Domain fine-tuning benefits Phi-4 and Whisper but degrades MMS, revealing limitations of adapter-based adaptation. Proxy language adapters outperform training MMS adapters from scratch.
Distributional metrics such as Fréchet Audio Distance cannot score individual music clips and correlate poorly with human judgments, while the only per-sample learned metric achieving high human correlation is closed-source. We introduce MUQ-EVAL, an open-source per-sample quality metric for AIgenerated music built by training lightweight prediction heads on frozen MuQ-310M features using MusicEval, a dataset of generated clips from 31 text-to-music systems with expert quality ratings. Our simplest model, frozen features with attention pooling and a two-layer MLP, achieves system-level SRCC = 0.957 and utterance-level SRCC = 0.838 with human mean opinion scores. A systematic ablation over training objectives and adaptation strategies shows that no addition meaningfully improves the frozen baseline, indicating that frozen MuQ representations already capture quality-relevant information. Encoder choice is the dominant design factor, outweighing all architectural and training decisions. LoRA-adapted models trained on as few as 150 clips already achieve usable correlation, enabling personalized quality evaluators from individual listener annotations. A controlled degradation analysis reveals selective sensitivity to signal-level artifacts but insensitivity to musical-structural distortions. Our metric, MUQ-EVAL, is fully open-source, outperforms existing open per-sample metrics, and runs in real time on a single consumer GPU. Code, model weights, and evaluation scripts are available at https://github.com/dgtql/MuQ-Eval.
Software synthesizers expose dozens of coupled controls, many of them redundant or perceptually inert at typical operating points. This paper presents a two-stage, measure-and-validate framework that ranks parameters by their influence on the audio output and then checks that ranking against an independent, variance-based experiment. Every rendered clip is normalized to -23 LUFS and described by five spectral fingerprints, so that gain shifts cannot masquerade as timbral change. The Sweep Test scores each parameter by spectral deviation over a 101-step sweep; the Sobol Test scores it by summed per-cell variance over a 128-point Sobol excitation. Across five architecturally distinct synthesizers (TorchSynth, Pyo, Pedalboard, Csound, and FluidSynth) spanning 48 to 76 parameters, the two tests agree at Spearman ρ= 0.82-0.96 (p-values \le 7.2 \times 10^{-15}) on the four engines that render deterministically, and at ρ= 0.32 on the one with process-level render noise--a diagnostic of that engine rather than of the method. The strong correlations survive reducing the Sobol budget from 128 to 32 points, and the complete analysis costs a one-time offline budget of about 66,000 renders.
Hankun Wang, Bohan Li, Shi Lian, Xiaoyu Gu, Jing Peng, Da Zheng, Yiwei Guo, Colin Zhang 等 10 位作者
Speech editing for content creation requires precise control over both what an edit should do and where it should apply. Free-form natural language provides a flexible interface for expressing edit requests, but its ambiguity may leave the intended operation, parameters, or target region underspecified. We study a semi-formal transcript-grounded edit instruction that specifies operation categories and parameters and binds them to transcript spans or boundaries. This semantic timeline avoids explicit timestamp alignment and provides an externally inspectable contract for compositional edits. We instantiate the interface in dots$.$tts$.$edit, a speech editing model built on a continuous autoregressive speech generation architecture. Four representative speech-creation controls cover lexical content, affective expression, pitch and speaking-rate delivery, and temporal phrasing through text, emotion, prosody, and pause editing. Task-specific data pipelines construct operation- and scope-controlled pairs while retaining source-derived context outside each target region. We further introduce doteBench, a bilingual benchmark that measures precise instruction following, local preservation, and audio quality across the four controls and their composition. Experiments show leading overall instruction following and local preservation across its five editing categories, while audio quality remains comparable to existing open-source systems. Across three Seed-TTS-Eval shards, the model shows negligible differences from the base model in zero-shot TTS recognition error rate and speaker similarity.
SpecAugment is the dominant speech augmentation policy, but researchers have studied it primarily for automatic speech recognition (ASR), and it operates on spectrograms, limiting its applicability to other speech tasks and waveform models. We propose Wav2Aug, a drop-in waveform augmentation policy that randomly samples two augmentations from a calibrated pool of waveform transformations. We calibrate the augmentation hyperparameters once and reuse them across tasks and architectures, eliminating task-specific tuning. Experiments on ten speech benchmarks show that Wav2Aug achieves ASR performance comparable to SpecAugment while improving over it on average across non-ASR tasks, yielding an average relative error reduction of 12.8% over the strongest compatible SpecAugment baseline. We further show that multi-view training provides additional gains, with two augmented views offering the best tradeoff between performance and computational cost.
Video-queried sound separation (VQSS) aims to isolate a target sound from an audio mixture using its corresponding video as a query. However, scene-level video queries can become ambiguous when multiple sounding objects are visible, particularly when the target and interferer are semantically similar. We propose SelectSep, a target-selective VQSS framework that jointly uses text and a single click coordinate to specify the target source and instance without requiring segmentation at inference time. We adapt the visual encoder through target-feature distillation with text and click conditioning, and then fine-tune the separator using the adapted visual features. We also introduce an on-screen interference setting, where both the target and interferer are visible, to evaluate target selectivity. Experiments on standard benchmarks and the on-screen interference setting demonstrate consistent improvements in separation quality and target selectivity, particularly for mixtures containing semantically similar sources.
Large audio language models (LALMs) can process speech directly, but it remains unclear how well they use the non-lexical acoustic information that speech carries. We study this question through pain, using TAME-Pain, which contains pain-neutral Harvard sentences and spoken pain ratings recorded at different self-reported pain levels. We introduce HEAR, a four-stage diagnostic framework that traces pain-related information from the speech signal to the model's prediction, and apply it to seven closed-source and open-weight LALMs. On the Harvard sentences, where pain can only be inferred from the voice, all models perform at chance level, and same-speaker references or few-shot examples improve balanced accuracy by no more than six points. On the spoken ratings, the same models reach 87.5-99.7% balanced accuracy. Layer-wise probing of open-weight LALMs shows that pain can be decoded from early audio representations (74.1% balanced accuracy for binary prediction and 45.0% for three-class severity), but that this information, along with most pain-associated acoustic cues, decreases in deeper layers. In an acoustic perception test, weaker models cannot distinguish even large changes in these cues, whereas stronger models can but still fail to infer pain. Finally, we recover pain-related acoustic cues and directly reinject them as structured text into the prompt, which improves pairwise pain prediction by up to 32 points. These results indicate that current LALMs encode paralinguistic information in early layers that is largely unavailable when the model produces its answer.
Diankun An, Lu Ma, Yongliang Wang, Zhewen Su, Xin Tian
This paper addresses fine-grained prosody control for large-language-model (LLM) based text-to-speech (TTS) synthesis of Mandarin Chinese, motivated by multimedia content-creation scenarios---film dubbing, audiobook production, and interactive voice interfaces---in which prosodic adjustments must be placed on specific characters rather than applied uniformly to whole utterances. We present a data-driven framework in which fine-grained prosodic markers (pitch, duration, energy, and pause) are automatically extracted from raw audio at scale and embedded into the text stream as out-of-vocabulary Unicode symbols. The LLM of an existing TTS system is then fine-tuned to interpret these markers while its architecture remains unchanged, yielding character-level temporal control that complements existing utterance-level style control. An optional text-LLM front-end can translate natural-language instructions into marker sequences. Experiments on 2.15M Mandarin utterances (5,858 hours) with CosyVoice2 as the backbone demonstrate reliable control of character duration, pitch, pause duration, and global speaking rate, with perceptual quality essentially preserved. A cross-tokenizer analysis further indicates that the ASR-optimized tokenizer of the backbone attenuates prosodic variation relative to reconstruction-oriented alternatives---an observation we report as hypothesis-generating evidence for prosody-sensitive system design.
Automatic speech recognition systems commonly rely on reference transcriptions for evaluation, while reference-free approaches often depend on internal confidence estimation or auxiliary language models. We propose READ (Reference-free Hypothesis Evaluation with Acoustic Discrepancy), a novel metric that evaluates ASR hypotheses directly from the speech signal. READ emphasizes the acoustic grounding of hypotheses. It uses a pretrained auto-regressive TTS model to compute the conditional likelihood of speech tokens given a text hypothesis, to measure fine-grained acoustic discrepancy between speech and text. Without additional training, READ can be applied for hypothesis refinement. Experiments show that READ correlates with specific recognition errors and improves ASR outputs, achieving up to 20\% relative error rate reduction, with particularly strong gains under noisy conditions.
Sripathi Sridhar, Yujin Kim, Giovana Morais, Liang-Yuan Wu, Mehmet Atilay Kucukoglu, Prem Seetharaman, Lloyd May, Mark Cartwright 等 9 位作者
Audio captioning benchmarks typically evaluate a single unstructured free-text description of an audio scene, providing limited insight into how a model fails to understand polyphonic audio. A plausible caption may omit a source, or conflate multiple sources into one, or assign an attribute of one source to another, and conventional captioning evaluation cannot locate these errors. We introduce MixCardBench, a benchmark for structured captioning of polyphonic audio, which evaluates models in two stages: first, identifying and temporally localizing the sources in a scene, and second, generating a structured MixCard describing the semantic, temporal, and acoustic properties of each source. The benchmark contains 4884 ecologically plausible synthetic soundscapes with up to nine overlapping speech, music and sound effects sources built from high-quality recordings and source-level annotations, where each source has a reference MixCard. We evaluate six state-of-the-art large audio language models and find substantial limitations in current models, which struggle considerably with speech and sound effects timing and detection, with the strongest model reaching only around 60% mean card-field score. Performance varies significantly across attributes, exposing capability differences that are hidden by aggregate caption scores. We assess the quality and ecological validity of MixCardBench through human validation, distribution matching and a listening study, and show that human corrections have little effect on model rankings, and our synthesized soundscapes are closer to in-the-wild YouTube videos than source guided but otherwise random synthesis. The benchmark dataset and code are publicly available, and demo examples are at https://mixcards-anon.github.io/mixcards-demo/
Music audio generation spans instrumental clips, full songs with vocals, edits of existing recordings, and systems that play alongside a musician in real time. This monograph explains the methods behind these systems for readers familiar with neural-network training who are new to music generation. It introduces neural audio codecs and autoencoders, autoregressive and masked token models, diffusion and flow matching, sampling acceleration, and preference optimization in a shared notation. Worked examples connect training objectives and sampling procedures to published systems. The application sections relate the musical requirements of instrumental generation, songs, editing, and live performance to model design. Comparison tables record representations, methods, inputs, tasks, and release information documented in papers and official sources. Symbolic composition, speech, and sound effects are included where they connect to music audio. We explain how to interpret automatic metrics and learned judges, design and report listening tests, and assess evidence about data provenance, memorization, attribution, and output detection. Selected legal decisions and disputes illustrate the implications for training and release. The reviewed systems combine established generative paradigms with new representations, planning stages, and post-training procedures. Differences in data, scale, and evaluation limit what cross-system comparisons can establish about the sources of reported gains.
Maris Basha, Yuhang Wang, Xiaoran Chen, Longbiao Cheng, Luca Yapura, Anja T. Zai, Mathieu Salzmann, Richard Hahnloser
Deciding which member of an animal group produced a vocalization and where that vocalization was produced are central to studying social communication but are rarely evaluated with physically verified ground truth. Existing resources either localize sounds without identifying the caller, or recognize individuals from isolated vocalizations without reliance on the 3D candidate geometry. We close this gap with V33DA, a dataset-centered multimodal benchmark for spatial vocal attribution in social zebra finches: it comprises 33,625 verified events coupling 5-channel audio with three camera views, calibrated 3D pose for every visible candidate, and per-bird radio telemetry across 10 individuals and three experiments. Caller identity is verified from an on-body accelerometer-derived vibration signal; this channel is withheld from benchmark models and used only by an on-body reference ceiling. Given a detected vocalization and the visible candidate birds, the task is to select the caller. We evaluate it under three regimes: session-disjoint testing on familiar birds from held-out recording dates, held-out-experiment transfer to a new cohort, and leave-one-individual-out evaluation on a bird whose vocalizations are held out from training. Reference methods expose a consistent gap between in-domain accuracy and transfer to unseen individuals: methods that can exploit familiar vocal identity perform well on known callers but degrade sharply on unseen ones, whereas candidate-aware methods preserving explicit spatial reasoning transfer substantially better. The on-body reference stays near-perfect across regimes, confirming the internal consistency of the accelerometer-verified labels. We release V33DA with fixed protocols and code for the evaluated methods, and V33DA++, an unevaluated companion set of verified overlapping-call events and +/-2 s context windows.
William Chen, Prem Seetharaman, Ke Chen, Oriol Nieto, Kevin Duarte, Siddharth Srinivasan Iyer, Mamshad Nayeem Rizve, Zhiwen Cao 等 13 位作者
Multimedia editing requires generative models to interpret complex, compositional instructions and modify only the desired elements across modalities, a capability largely beyond existing systems. Current approaches are typically limited to simple, single-attribute edits within one modality, since diverse, high-quality editing pairs are hard to obtain, especially for cross-modal tasks where aligned audiovisual (AV) data is scarce. We present CrossEdit, a unified omni-modal editing model for images, audio, and video that performs AV movie scene edits zero-shot through cross-modal transfer. Our key observation is that complex instruction-following, once learned in any modality, generalizes to others. We exploit the additive nature of acoustic signals to procedurally generate large-scale audio editing pairs with compositional instructions, and show that fine-tuning on this synthetic data and self-supervised AV masked reconstruction, alongside a targeted set of curated cross-modal tasks, induces instruction-following that transfers zero-shot to unseen modality and instruction combinations. To evaluate this capability, we release CrossEditBench, a human-annotated benchmark of AV edits on movie scenes, and propose AV-FES, a metric that jointly scores instruction following and consistency. We show that our proposed techniques improve performance on zero-shot AV movie scene editing, while maintaining or improving performance on lip-synced speech editing and standard image, video, and audio editing benchmarks. Demos are available at https://wanchichen.github.io/crossedit/.
Jinzuomu Zhong, Alice Ross, Catherine Lai, Korin Richmond
Utterance-level similarity ratings do not reveal where perceived accent differences occur in generated speech. We introduce a segmental evaluation protocol combining speaker and accent similarity ratings with subword annotations of perceived accent differences provided by listeners from target accent regions. Using newly collected recordings, 47 retained listeners evaluate 200 content-matched reference--candidate pairs spanning four Scottish and Australian voices, two zero-shot TTS systems, and an articulation-based voice conversion system. Subword annotations locate salient pronunciation differences and sometimes yield different system rankings from similarity ratings. Speaker and accent ratings remain strongly associated ($ρ= 0.71$), highlighting the value of examining local perceptual evidence beyond these scores. Evaluation of seven objective distances reveals weak correspondence with subword selections, with correlations reaching at most $ρ= 0.22$. We release code, data, and an interactive demo to support the development of perceptually grounded segmental accent evaluation.
Ioannis Stylianou, Achintya kr. Sarkar, Nauman Dawalatabad, James Glass, Zheng-Hua Tan
Robust Voice Activity Detection (VAD) remains a challenging task, especially under noisy, diverse, and unseen acoustic conditions. Beyond algorithmic development, a key limitation in advancing VAD research is the lack of large-scale, systematically controlled, and publicly available datasets. To address this, we introduce \textbf{LibriVAD} -- a scalable open-source dataset derived from LibriSpeech and augmented with diverse real-world and synthetic noise sources. LibriVAD enables systematic control over speech-to-noise ratio, silence-to-speech ratio (SSR), and noise diversity, and is released in three sizes (15 GB, 150 GB, and 1.5 TB) with two variants (LibriVAD-NonConcat and LibriVAD-Concat) to support different experimental setups. We benchmark multiple feature-model combinations, including waveform, Mel-Frequency Cepstral Coefficients (MFCC), and Gammatone filter bank cepstral coefficients, and introduce the Vision Transformer (ViT) architecture for VAD. Our experiments show that ViT with MFCC features consistently outperforms established VAD models such as boosted deep neural network and convolutional long short-term memory deep neural network across seen, unseen, and out-of-distribution (OOD) conditions, including evaluation on the real-world VOiCES dataset. We further analyze the impact of dataset size and SSR on model generalization. Experimental results show that scaling the dataset size from small-scale to medium- or large-scale noticeably and consistently improves VAD performance under OOD conditions, while appropriately balancing SSR enhances OOD performance in the small-scale setting. All datasets, trained models, and code are publicly available to foster reproducibility and accelerate progress in VAD research.
Manan Mittal, Kanad Sarkar, Ryan M. Corey, John R. Buck, Andrew C. Singer
The short-time Fourier transform (STFT) characterizes the local frequency content of time-varying signals by evaluating the discrete Fourier transform over a sliding analysis window. The STFT is governed by a fundamental resolution trade-off: a short window localizes rapid transients in time at the cost of frequency resolution, whereas a long window resolves closely spaced harmonics at the expense of temporal resolution. The Universal Short-Time Fourier Transform (U-STFT) resolves this trade-off sequentially by modeling the analysis window duration as a data-adaptive state variable that indexes the number of hops included in the current observation interval. At each frame, a bank of candidate window durations is scored using normalized order-three Rényi entropy to quantify joint time-frequency concentration. A window length trellis aggregates the resulting $\mathcal{O}(3^M)$ candidate window schedules with linear complexity per frame, guaranteeing a cumulative mix-loss regret bound of $\mathcal{O}((k+1)\log M)$ relative to any hindsight window schedule with $k$ resolution changes over $M$ frames. By linearity of the Fourier transform, posterior expectation across the candidate bank yields a single spectrum equivalent to an STFT evaluated with a data-adaptive effective window, while an envelope-normalized synthesis stream guarantees exact overlap-add reconstruction. Evaluations on synthetic and marine bioacoustic signals demonstrate adaptive resolution tracking, enhanced spectral concentration, and exact time-domain invertibility.