Daily Papers

Spoken interaction is moving from alternating turns to full-duplex conversation, where an agent listens while speaking. Learning this coordination requires natural dialogue with persistent speaker-resolved tracks, a data requirement that remains difficult to satisfy at scale. Generative recovery offers a route from ordinary recordings, but recovering two voices is not enough: their track assignments can drift as the conversation unfolds. We address this missing link with speaker-anchored routing before diffusion: a 1.84M-parameter module matches the separator's coarse streams to fixed enrollment identities and orders them before a frozen generator reconstructs the speech, without retraining the generator or altering the reconstructed audio. On 25-minute remote conversations, routing cuts wrong-track time from 43.9% to 4.1%, below the 6.2% of matched-enrollment waveform re-identification, with its largest advantage when one participant holds the floor (17.8% versus 28.6%), and raises fixed-track activity F1 from 0.46 to 0.84 while per-chunk reconstruction quality is unchanged. The same router transfers to unseen face-to-face and telephone recordings, and a margin-gated verification step brings their wrong-track time to 0.9-1.7% while embedding 20-36% of the output, where full re-identification embeds all of it. Recovered voices thus become the participant-consistent tracks that full-duplex dialogue modeling needs. Audio examples: https://elpsykongloo.github.io/beyond-two-tracks/

In recent years, Large Language Models (LLMs) have enabled users to provide highly specific music recommendation requests using natural language prompts (e.g. "Can you recommend some old classics for slow dancing?"). In this setup, the recommended tracks are predicted by the LLM in an autoregressive way, i.e. the LLM generates the track titles one token at a time. While intuitive, this approach has several limitation. First, it is based on a general purpose tokenization that is optimized for words rather than for track titles. Second, it necessitates an additional entity resolution layer that matches the track title to the actual track identifier. Third, the number of decoding steps scales linearly with the length of the track title, slowing down inference. In this paper, we propose to address the task of prompt-based music recommendation as a generative retrieval task. Within this setting, we introduce novel, effective, and efficient representations of track identifiers that significantly outperform commonly used strategies. We introduce Text2Tracks, a generative retrieval model that learns a mapping from a user's music recommendation prompt to the relevant track IDs directly. Through an offline evaluation on a dataset of playlists with language inputs, we find that (1) the strategy to create IDs for music tracks is the most important factor for the effectiveness of Text2Tracks and semantic IDs significantly outperform commonly used strategies that rely on song titles as identifiers (2) provided with the right choice of track identifiers, Text2Tracks outperforms sparse and dense retrieval solutions trained to retrieve tracks from language prompts.

We study timestamped speaker-attributed automatic speech recognition (SA-ASR) for long-form, multi-party speech with overlap. In this setting, chunk-wise inference must preserve meeting-level speaker identity consistency while producing time-stamped, speaker-labeled transcripts. Prior Speech-LLM systems tend to prioritize either local diarization or global labeling, lacking the ability to jointly model fine-grained temporal boundaries and robust cross-chunk identity linking. We propose G-STAR, an end-to-end framework that couples a cache-conditioned speaker-tracking module with a Speech-LLM transcription backbone. The tracker provides structured speaker cues with temporal grounding, and the LLM generates attributed text conditioned on these cues. G-STAR supports component-wise optimization and joint end-to-end training, enabling flexible learning under heterogeneous supervision and domain shift. Under chunk-wise decoding protocols, experiments on both oracle-segmented local evaluation and full-meeting global evaluation show strong speaker-attributed transcription performance.

Speech-guided referring video object segmentation aims to recover the mask tracks of objects specified by a spoken motion description. Here, speech carries a linguistic instruction rather than acoustic evidence from a sounding object, so a solution must connect speech recognition, motion-centric temporal grounding, mask tracking, and explicit no-target handling. We introduce Speech2MaskTrack, our approach for the MeViS-Audio track of the 8th LSVOS Challenge. Speech2MaskTrack transcribes the spoken query and compiles it into structured constraints over category, count, direction, interaction role, and temporal phase. SAM3.1 enumerates multiple instance tracks, which TRACE ranks using complete-trajectory motion and relation evidence. A frozen lexical presence gate may suppress the ranked SAM3.1 base prediction. When the gate predicts that a target is present, an available full-expression-conditioned SaSaSa2VA track replaces the SAM3.1 mask. Only outputs that remain empty enter GPT-assisted recovery, which invokes SaSaSa2VA again under query- and mask-level verification. Speech2MaskTrack achieved second place in the official challenge ranking.

Speech language models (speech LLMs) can generate plausible outputs from audio that contains no usable speech evidence. We study this failure as a pre-generation support-estimation problem and present SURE-Voice, a training-free front end that decides whether an audio prompt contains intelligible speech evidence before calling a speech LLM. We build SURE-Challenge with a 640-example SURE-Core split and a 1,920-example SURE-Extended split derived from 120 LibriSpeech source utterances. Using one fixed operating point, an energy screen plus Whisper token confidence raises unsupported accuracy on the held-out Extended test from 0.000--0.133 to 0.919 for six non-degenerate speech LLM backbones, while supported accuracy remains 0.919--0.970 and downstream calls fall from 480 to 287. A 500-clip ESC-50 sanity set shows the same pattern on real environmental audio, with vocal non-speech as a residual failure mode. An overlap diagnostic shows that source attribution remains separate from speech-evidence filtering. The evidence supports a controlled benchmark baseline and a deployment-oriented analysis; it does not establish universal robustness to semantic answerability, gain variation or natural conversations.

Probing is widely adopted in computer vision to faithfully evaluate self-supervised learning (SSL) embeddings, as finetuning may misrepresent their inherent quality. In contrast, audio SSL models still rely on finetuning because simple probing fails to unlock their full potential and alters their rankings when competing on AudioSet. Hence, a robust and efficient probing mechanism is required to guide the trajectory of audio SSL towards reliable and reproducible methods. We introduce Convex Gated Probing (CGP), a prototype-based method that significantly closes the gap between finetuning and probing in audio. CGP efficiently utilizes all frozen layers via a gating mechanism and exposes the location of latent task-relevant information. Guided by CGP as a reliable post-hoc evaluation probe, we rework the entire SSL pipeline of current best performing audio models that use legacy implementations of prior SSL methods. By refining data preprocessing, model architecture, and pretraining recipe, we introduce Better Audio Transformer (BAT), and establish new SOTA on audio benchmarks.

Speech retrieval has become increasingly important as spoken content continues to grow across meetings, lectures, podcasts, and videos. Existing benchmarks and models have advanced semantic search over spoken content, but largely focus on what is said while overlooking who says it. In many real-world scenarios, however, users need to retrieve speech based jointly on semantic content and a target speaker, where the speaker may be specified naturally through a reference speech utterance rather than a predefined identity. To address this gap, we introduce VoiceTrace-Bench, a benchmark for hybrid speech retrieval in which each query combines text specifying what to retrieve with reference speech specifying who to retrieve. This setting requires models to integrate complementary semantic and speaker information directly from heterogeneous query inputs. Motivated by the joint audio-text modeling capabilities of audio-language models (ALMs), we develop VoiceTrace, a two-stage retrieval framework consisting of VoiceTrace-Emb, an embedding model that learns unified representations for efficient large-scale retrieval, and VoiceTrace-Reranker, a reranking model that jointly examines each query--candidate pair for fine-grained relevance estimation. Experiments show that VoiceTrace achieves state-of-the-art performance on established semantic speech retrieval benchmarks, while substantially outperforming cascade-based approaches on VoiceTrace-Bench, demonstrating its effectiveness for both conventional semantic retrieval and the new hybrid retrieval setting.

Audio super-resolution is challenging owing to its ill-posed nature. Recently, the application of diffusion models in audio super-resolution has shown promising results in alleviating this challenge. However, diffusion-based models have limitations, primarily the necessity for numerous sampling steps, which causes significantly increased latency when synthesizing high-quality audio samples. In this paper, we propose FLowHigh, a novel approach that integrates flow matching, a highly efficient generative model, into audio super-resolution. We also explore probability paths specially tailored for audio super-resolution, which effectively capture high-resolution audio distributions, thereby enhancing reconstruction quality. The proposed method generates high-fidelity, high-resolution audio through a single-step sampling process across various input sampling rates. The experimental results on the VCTK benchmark dataset demonstrate that FLowHigh achieves state-of-the-art performance in audio super-resolution, as evaluated by log-spectral distance and ViSQOL while maintaining computational efficiency with only a single-step sampling process.

Sound event localization and detection (SELD) combines the identification of sound events with the corresponding directions of arrival (DOA). Recently, event-oriented track output formats have been adopted to solve this problem; however, they still have limited generalization toward real-world problems in an unknown polyphony environment. To address the issue, we proposed an angular-distance-based multiple SELD (AD-YOLO), which is an adaptation of the "You Only Look Once" algorithm for SELD. The AD-YOLO format allows the model to learn sound occurrences location-sensitively by assigning class responsibility to DOA predictions. Hence, the format enables the model to handle the polyphony problem, regardless of the number of sound overlaps. We evaluated AD-YOLO on DCASE 2020-2022 challenge Task 3 datasets using four SELD objective metrics. The experimental results show that AD-YOLO achieved outstanding performance overall and also accomplished robustness in class-homogeneous polyphony environments.

Recent audio-video generators can produce semantically plausible and apparently synchronized sound, yet may still violate the acoustic processes implied by visible events and environments. Existing benchmarks provide limited support for attributing such violations to particular acoustic processes and quantifying their severity. We introduce AcoustiTrace, a diagnostic benchmark that formalizes acoustic physical realism in audio-video generation. AcoustiTrace organizes text-to-audio-video (T2AV) and image-to-audio-video (I2AV) evaluation around the acoustic process, covering sound generation, propagation environment, and acoustic reception through eight dimensions grounded in measurable acoustic quantities. Based on these evaluation dimensions, we construct a large-scale dataset organized around acoustic mechanisms, comprising real-world audio-video recordings and acoustically annotated RGB-D observations, and use it to develop targeted prompt suites and validated evaluators. Experiments reveal that even leading generators still struggle with fundamental acoustic processes despite producing plausible sound events. Finally, we show that the diagnostics AcoustiTrace provides for specific acoustic relations can guide model refinement toward more physically faithful audio and open new directions for incorporating acoustic principles into training objectives, reward modeling, and candidate selection.

Latest advances in deep spatial filtering for Ambisonics demonstrate strong performance in stationary multi-speaker scenarios by rotating the sound field toward a target speaker prior to multi-channel enhancement. For applicability in dynamic acoustic conditions with moving speakers, we propose to automate this rotary steering using an interleaved tracking algorithm conditioned on the target's initial direction. However, for nearby or crossing speakers, robust tracking becomes difficult and spatial cues less effective for enhancement. By incorporating the processed recording as additional guide into both algorithms, our novel joint autoregressive framework leverages temporal-spectral correlations of speech to resolve spatially challenging speaker constellations. Consequently, our proposed method significantly improves tracking and enhancement of closely spaced speakers, consistently outperforming comparable non-autoregressive methods on a synthetic dataset. Real-world recordings complement these findings in complex scenarios with multiple speaker crossings and varying speaker-to-array distances.

Recent Large Audio-Language Models (LALMs) exhibit impressive capabilities in understanding audio content for conversational QA tasks. However, these models struggle to accurately understand timestamps for temporal localization (e.g., Temporal Audio Grounding) and are restricted to short audio perception, leading to constrained capabilities on fine-grained tasks. We identify three key aspects that limit their temporal localization and long audio understanding: (i) timestamp representation, (ii) architecture, and (iii) data. To address this, we introduce TimeAudio, a novel method that empowers LALMs to connect their understanding of audio content with precise temporal perception. Specifically, we incorporate unique temporal markers to improve time-sensitive reasoning and apply an absolute time-aware encoding that explicitly grounds the acoustic features with absolute time information. Moreover, to achieve end-to-end long audio understanding, we introduce a segment-level token merging module to substantially reduce audio token redundancy and enhance the efficiency of information extraction. Due to the lack of suitable datasets and evaluation metrics, we consolidate existing audio datasets into a new dataset focused on temporal tasks and establish a series of metrics to evaluate the fine-grained performance. Evaluations show strong performance across a variety of fine-grained tasks, such as dense captioning, temporal grounding, and timeline speech summarization, demonstrating TimeAudio's robust temporal localization and reasoning capabilities.

Speech transcripts used as long-term memory must preserve both words and stable speaker identities. Existing meeting-transcription metrics either ignore speakers or remap anonymous speakers independently in each recording, so they cannot measure whether the same person retains one identity across meetings. We evaluate persistent speaker attribution with Speaker Identified cpWER (SI-cpWER), which scores a corpus under one global speaker-ID assignment. The benchmark covers five commercial diarize-then-identify cascades, two open academic baselines, and ThyVoice on the full 129-meeting CHiME-8 NOTSOFAR evaluation set in clean and noiseaugmented form, plus CHiME-6. ThyVoice is our end-to-end reference system; it repairs overlap and gates the evidence used to create and update voiceprints. Requiring persistent identity changes the commercial ranking: ThyVoice records lower SI-cpWER than every evaluated commercial cascade in all three conditions and the lowest mean in the full panel, 47.13 versus 54.75 for the next system. Complementary lexical, diarization, per-recording attribution, and speaker-clustering diagnostics characterize upstream error surfaces in the final attributed record. These results show why persistent attribution must be evaluated directly in systems that reuse conversations across time.

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.

Full-length song generation must preserve coherence and musicality, render detailed vocal and accompaniment acoustics, and follow lyrics and prompts. Existing language model-based systems face a structural trade-off: mixed-token modeling preserves vocal-instrument coordination but obscures track-specific details, whereas dual-track prediction improves acoustics but requires longer sequences and weakens global planning. We present LeVo 2, a hybrid LLM-Diffusion framework for controllable full-length song generation. LeVo 2 formulates this trade-off as hierarchical modeling: LeLM first predicts mixed tokens for semantic planning, then predicts vocal and accompaniment tokens in parallel for track-specific refinement, while a diffusion-based Music Codec reconstructs full-length waveforms. A central contribution of this extended version is an aesthetics-guided training schedule for alignment. During pre-training, an automated music aesthetic evaluation framework assigns musicality-tier conditions to large-scale data, providing musicality priors before preference alignment. Progressive post-training applies SFT, large-scale offline DPO, and closed-loop semi-online DPO to separately improve generation quality, controllability, and musicality. Modular extension then trains the Track-Specific LM for acoustic refinement while preserving the aligned semantic planner. This schedule separates musicality learning, controllability alignment, and acoustic refinement, mitigating optimization conflict and the limitations of static offline preference pairs. Expert listening tests and objective evaluations show that LeVo 2 outperforms open-source baselines across six subjective dimensions, and approaches leading commercial systems on several listening metrics. Ablations validate the effects of the training strategy, aesthetics guidance, scaling, and hierarchical architecture.

Transcribing and understanding multi-speaker conversations requires speech recognition, speaker attribution, and timestamp localization. While speech LLMs excel at single-speaker tasks, multi-speaker scenarios remain challenging due to overlapping speech, backchannels, rapid turn-taking, and context window constraints. We propose Speaker-Reasoner, an end-to-end Speech LLM with agentic multi-turn temporal reasoning. Instead of single-pass inference, the model iteratively analyzes global audio structure, autonomously predicts temporal boundaries, and performs fine-grained segment analysis, jointly modeling speaker identity, gender, timestamps, and transcription. A speaker-aware cache further extends processing to audio exceeding the training context window. Trained with a three-stage progressive strategy, Speaker-Reasoner achieves consistent improvements over strong baselines on AliMeeting and AISHELL-4 datasets, particularly in handling overlapping speech and complex turn-taking.

Deep Speech Enhancement Challenge is the 5th edition of deep noise suppression (DNS) challenges organized at ICASSP 2023 Signal Processing Grand Challenges. DNS challenges were organized during 2019-2023 to stimulate research in deep speech enhancement (DSE). Previous DNS challenges were organized at INTERSPEECH 2020, ICASSP 2021, INTERSPEECH 2021, and ICASSP 2022. From prior editions, we learnt that improving signal quality (SIG) is challenging particularly in presence of simultaneously active interfering talkers and noise. This challenge aims to develop models for joint denosing, dereverberation and suppression of interfering talkers. When primary talker wears a headphone, certain acoustic properties of their speech such as direct-to-reverberation (DRR), signal to noise ratio (SNR) etc. make it possible to suppress neighboring talkers even without enrollment data for primary talker. This motivated us to create two tracks for this challenge: (i) Track-1 Headset; (ii) Track-2 Speakerphone. Both tracks has fullband (48kHz) training data and testset, and each testclips has a corresponding enrollment data (10-30s duration) for primary talker. Each track invited submissions of personalized and non-personalized models all of which are evaluated through same subjective evaluation. Most models submitted to challenge were personalized models, same team is winner in both tracks where the best models has improvement of 0.145 and 0.141 in challenge's Score as compared to noisy blind testset.

A common design pattern in high-quality music generation is to handle structure and fidelity in different representation spaces: a generator first models high-level structure, followed by diffusion-based or neural decoding stages that reconstruct fine details. In this work, we explore an alternative view: both may be progressively modeled within a single deep acoustic-token hierarchy. To study this, we build a 64-layer residual vector quantization (RVQ) acoustic representation and propose a two-stage coarse-to-fine generation framework. A backbone model first generates coarse acoustic tokens for the full track, and a super-resolution model then completes finer tokens within the same acoustic token space. The super-resolution stage works at full-track scale and refines tokens layer by layer while running in parallel over time, leading to a fixed 62-step inference process. To jointly improve lyric alignment and fine-detail reconstruction, we further introduce hybrid-attention training: the alignment objective uses causal attention, while layer-wise refinement uses full attention. A key finding is that text--vocal alignment can emerge within pure acoustic-token language modeling, without requiring a separate semantic token stage. Moreover, initializing the super-resolution model from the trained backbone significantly improves convergence and final quality. Taken together, our results suggest that high-quality music generation can be effectively pursued without separating structure and fidelity into heterogeneous representation spaces. Instead, both can be progressively modeled within a unified acoustic-token hierarchy, pointing toward a simpler and more unified path to high-quality music generation.

Zero-shot text-to-speech (TTS) now supports interactive assistants, personalized media, and accessibility tools. All TTS systems require faithful linguistic rendering, consistent speaker identity, and low-latency response. Yet compact streaming systems must preserve sufficient acoustic detail in a predictable low-rate latent sequence, while iterative diffusion sampling and classifier-free guidance multiply inference cost at every autoregressive step. To strike a balance between high-fidelity synthesis and low-latency inference, we present CuteTTS, a compact continuous-autoregressive TTS system. It combines semantically aligned causal VAE latents with patch-level autoregression, explicit speaker conditioning, and a bidirectional flow-matching head. We further introduce guidance-step distillation, which absorbs classifier-free guidance and multiple solver steps into a single interval-conditioned student. Evaluations on LibriSpeech and Seed-TTS-Eval demonstrate competitive intelligibility and speaker similarity in zero-shot voice cloning, while distillation lowers first-audio latency by 23.3% and real-time factor by 40.8% relative to the base model with comparable objective and subjective quality. These results provide a practical path toward continuous-autoregressive TTS that reconciles high-fidelity generation with the latency demands of real-time interaction.

While recent work in controllable text-to-audio (TTA) generation has achieved fine-grained control through timestamp conditioning, its scope remains limited by audio quality and input format. These models often suffer from poor audio quality in real datasets due to sole reliance on synthetic data. Moreover, some models are constrained to a closed vocabulary of sound events, preventing them from controlling audio generation for open-ended, free-text queries. This paper introduces PicoAudio2, a framework that advances temporal-controllable TTA by mitigating these data and architectural limitations. Specifically, we use a grounding model to annotate event timestamps of real audio-text datasets to curate temporally-strong real data, in addition to simulation data from existing works. The model is trained on the combination of real and simulation data. Moreover, we propose an enhanced architecture that integrates the fine-grained information from a timestamp matrix with coarse-grained free-text input. Experiments show that PicoAudio2 exhibits superior performance in terms of temporal controllability and audio quality.

We study device-addressed speech detection under pre-ASR edge deployment constraints, where systems must decide whether to forward audio before transcription under strict latency and compute limits. We show that, in multi-speaker environments with temporally ambiguous utterances, this task is more effectively modelled as a sequential routing problem over interaction history than as an utterance-local classification task. We formalize this as Sequential Device-Addressed Routing (SDAR) and present the Selective Attention System (SAS), an on-device implementation that instantiates this formulation. On a held-out 60-hour multi-speaker English test set, the primary audio-only configuration achieves F1=0.86 (precision=0.89, recall=0.83); with an optional camera, audio+video fusion raises F1 to 0.95 (precision=0.97, recall=0.93). Removing causal interaction history (Stage~3) reduced F1 from 0.95 to 0.57+/-0.03 in the audio+video configuration under our evaluation protocol. Among the tested components, this was the largest observed ablation effect, indicating that short-horizon interaction history carries substantial decision-relevant information in the evaluated setting. SAS runs fully on-device on ARM Cortex-A class hardware (<150 ms latency, <20 MB footprint). All results are from internal evaluation on a proprietary dataset evaluated primarily in English; a 5-hour evaluation subset may be shared for independent verification (Section 8.8).

Automatic speech recognition produces words but drops the non-verbal content of a recording: the breaths, coughs, sighs and laughter of the speaker, and the horns, music or barking around them. We describe Speech Event Detector, a 5.8-million-parameter convolutional-Transformer model that outputs probabilities every 40 ms for 54 event classes (20 speaker vocalizations and 34 background sounds), together with post-processing that converts these probabilities into timestamped events and a procedure for merging the events into a Whisper transcript. The model is trained on material drawn from seven public datasets whose labels come in incompatible forms: single detected regions, free-text tags with timestamps, temporally strong ontology labels, symbols embedded in transcripts, isolated recordings and environmental recordings. We unify them using rule-based label mapping, recover timestamps for transcript-embedded labels through CTC forced alignment, locating 6,360 of 8,620 sounds, and combine about 168 hours of real annotated training recordings with synthetic mixtures and augmentation using real background noise. Per-class thresholds for two operating profiles are tuned on real held-out recordings, and performance is estimated using two-fold cross-validation. On whole held-out recordings from four datasets, considering the 48 classes with at least ten held-out events, the sensitive profile detects an average of 49 percent of a class's events, covering 67 percent of all 9,367 events, at a macro precision of 0.30. The balanced profile achieves a macro recall of 0.37 at a precision of 0.36, with macro event F1 scores of 0.32 and 0.33, respectively. The detector runs about 39 times faster than real time on eight CPU threads.

Following the success of the 1st, 2nd, 3rd, 4th and 5th CHiME challenges we organize the 6th CHiME Speech Separation and Recognition Challenge (CHiME-6). The new challenge revisits the previous CHiME-5 challenge and further considers the problem of distant multi-microphone conversational speech diarization and recognition in everyday home environments. Speech material is the same as the previous CHiME-5 recordings except for accurate array synchronization. The material was elicited using a dinner party scenario with efforts taken to capture data that is representative of natural conversational speech. This paper provides a baseline description of the CHiME-6 challenge for both segmented multispeaker speech recognition (Track 1) and unsegmented multispeaker speech recognition (Track 2). Of note, Track 2 is the first challenge activity in the community to tackle an unsegmented multispeaker speech recognition scenario with a complete set of reproducible open source baselines providing speech enhancement, speaker diarization, and speech recognition modules.

Large Audio-Language Models (ALMs) have recently demonstrated remarkable capabilities in holistic audio understanding, yet they remain unreliable for temporal grounding, i.e., the task of pinpointing exactly when an event occurs within long-form audio. This limitation stems from two factors: training data dominated by clip-level supervision lacking precise timestamps, and benchmarks that fail to simulate real-world scenarios where short events are obscured by dense background sounds. In this paper, we introduce SpotSound, an audio language model designed for grounding audio events. SpotSound incorporates a novel training objective, specifically designed to suppress hallucinated timestamps for events absent from the input. Additionally, we present SpotSound-Bench, a challenging temporal grounding benchmark where target events occupy less than ~10\% of each clip, creating a rigorous `needle-in-a-haystack' evaluation. Experiments demonstrate that SpotSound achieves state-of-the-art results on temporal grounding benchmarks while maintaining robust performance across general downstream audio-language tasks. Code, models and benchmark are released on https://loiesun.github.io/spotsound/

In crowded settings, the human brain can focus on speech from a target speaker, given prior knowledge of how they sound. We introduce a novel intelligent hearable system that achieves this capability, enabling target speech hearing to ignore all interfering speech and noise, but the target speaker. A naive approach is to require a clean speech example to enroll the target speaker. This is however not well aligned with the hearable application domain since obtaining a clean example is challenging in real world scenarios, creating a unique user interface problem. We present the first enrollment interface where the wearer looks at the target speaker for a few seconds to capture a single, short, highly noisy, binaural example of the target speaker. This noisy example is used for enrollment and subsequent speech extraction in the presence of interfering speakers and noise. Our system achieves a signal quality improvement of 7.01 dB using less than 5 seconds of noisy enrollment audio and can process 8 ms of audio chunks in 6.24 ms on an embedded CPU. Our user studies demonstrate generalization to real-world static and mobile speakers in previously unseen indoor and outdoor multipath environments. Finally, our enrollment interface for noisy examples does not cause performance degradation compared to clean examples, while being convenient and user-friendly. Taking a step back, this paper takes an important step towards enhancing the human auditory perception with artificial intelligence. We provide code and data at: https://github.com/vb000/LookOnceToHear.

Traditional speaker-attributed ASR systems treated ASR and speaker diarization as two separate tasks. Recently, end-to-end models such as VibeVoice-ASR have unified the two tasks within a single model. However, existing unified models still mainly support offline recognition, making it difficult to meet the low-latency requirements of real-time voice assistants and agents. To tackle this issue, we present VibeVoice-ASR-Streaming, one of the first LLM-based end-to-end approaches to streaming speaker-attributed ASR. It interleaves fixed-size audio chunks, a small amount of lookahead audio and previous text. This allows the model to produce ''who said what'' as speech arrives, without a separate diarization stage. For transcription accuracy, our 7B model achieves the lowest average WER/CER across five evaluation sets. For speaker attribution, it achieves the best or tied-best on 12 of 13 evaluation settings. We release the 1.5B and 7B model weights together with inference code.

Speech restoration aims to remove distortions in speech signals. Prior methods mainly focus on single-task speech restoration (SSR), such as speech denoising or speech declipping. However, SSR systems only focus on one task and do not address the general speech restoration problem. In addition, previous SSR systems show limited performance in some speech restoration tasks such as speech super-resolution. To overcome those limitations, we propose a general speech restoration (GSR) task that attempts to remove multiple distortions simultaneously. Furthermore, we propose VoiceFixer, a generative framework to address the GSR task. VoiceFixer consists of an analysis stage and a synthesis stage to mimic the speech analysis and comprehension of the human auditory system. We employ a ResUNet to model the analysis stage and a neural vocoder to model the synthesis stage. We evaluate VoiceFixer with additive noise, room reverberation, low-resolution, and clipping distortions. Our baseline GSR model achieves a 0.499 higher mean opinion score (MOS) than the speech enhancement SSR model. VoiceFixer further surpasses the GSR baseline model on the MOS score by 0.256. Moreover, we observe that VoiceFixer generalizes well to severely degraded real speech recordings, indicating its potential in restoring old movies and historical speeches. The source code is available at https://github.com/haoheliu/voicefixer_main.

A unified audio model must recognize and understand linguistic, paralinguistic, and environmental information while supporting speech synthesis and editing. A key challenge is representation: understanding favors compact features suited to long-context modeling, whereas speech generation requires reconstructible features that preserve fine-grained acoustic detail. We introduce FireRedAudio, a general-purpose audio language model with a shared 9B-parameter LLM. To the best of our knowledge, it is the first publicly disclosed unified audio-language model to provide separate continuous input representations for understanding and generation within a single trainable autoregressive LLM. Audio to be recognized or analyzed is processed by a dedicated Audio Encoder, while speech inputs for generation use a RedAE-based pathway. The LLM directly generates text or conditions a flow-matching DiT to produce continuous acoustic latents. Through progressive multitask training, FireRedAudio supports ASR and audio understanding, with the latter extending to recordings of up to one hour, as well as zero-shot TTS, Instruct TTS, and semantic and acoustic speech editing. Its structured organization of long-form audio achieves second-level timestamp accuracy. Across comprehensive evaluations, FireRedAudio achieves competitive or leading performance in audio understanding and multilingual ASR, strong content accuracy and speaker preservation in zero-shot TTS, leading instruction following in Instruct TTS, and substantial improvements over Ming-UniAudio-Edit in both semantic and acoustic speech editing. These results demonstrate the viability of decoupled continuous input representations for unifying audio understanding and continuous-latent speech generation in a model of moderate scale. Our code is available at https://github.com/FireRedTeam/FireRedAudio.

Streaming text-to-speech is essential for low-latency spoken dialogue systems, yet many systems wait for sentence-level text and are therefore only pseudo-streaming. True token-level synthesis must generate speech from uncertain prefixes while maintaining perceptual continuity over an unbounded stream with bounded context. We present X2Streaming-TTS, a causal TTS framework that consumes asynchronously arriving text tokens and emits speech without accessing future input. To handle uncertain prefixes, we introduce causal commitment, which keeps ambiguous expressions provisional through uncertainty-aware buffering and performs capacity-adaptive, punctuation-aware segmentation. To preserve acoustic continuity, we further introduce causal speech-state inheritance, which carries the complete Code2Wav state and selected historical Talker states across segment boundaries. Together with an attention prior constraint, it blocks access to future positions while retaining bounded acoustic context. Experiments show that X2Streaming-TTS outperforms existing pseudo-streaming models on most subjective and objective metrics. Further analysis shows that causal commitment stabilizes online segmentation and reduces failures caused by insufficient context, while speech-state inheritance improves boundary continuity without degrading naturalness or speaker identity. X2Streaming-TTS thus achieves strict token-level synthesis with quality comparable to the evaluated offline baselines, a median time to first audio token (TTFT) of 15.8 ms for a single request, and a median TTFT of 260.8 ms at 128 concurrent requests. Our implementation is publicly available at https://github.com/X-Square-Robot/X2Streaming-TTS .

We propose a pre-training pipeline for audio spectrogram transformers for frame-level sound event detection tasks. On top of common pre-training steps, we add a meticulously designed training routine on AudioSet frame-level annotations. This includes a balanced sampler, aggressive data augmentation, and ensemble knowledge distillation. For five transformers, we obtain a substantial performance improvement over previously available checkpoints both on AudioSet frame-level predictions and on frame-level sound event detection downstream tasks, confirming our pipeline's effectiveness. We publish the resulting checkpoints that researchers can directly fine-tune to build high-performance models for sound event detection tasks.

Audio chaptering, the task of automatically segmenting long-form audio into coherent sections, is increasingly important for navigating podcasts, lectures, and videos. Despite its relevance, research remains limited and text-based, leaving key questions unresolved about leveraging audio information, handling ASR errors, and transcript-free evaluation. We address these gaps through three contributions: (1) a systematic comparison between text-based models with acoustic features, a novel audio-only architecture (AudioSeg) operating on learned audio representations, and multimodal LLMs; (2) empirical analysis of factors affecting performance, including transcript quality, acoustic features, duration, and speaker composition; and (3) formalized evaluation protocols contrasting transcript-dependent text-space protocols with transcript-invariant time-space protocols. Our experiments on YTSeg reveal that AudioSeg substantially outperforms text-based approaches, pauses provide the largest acoustic gains, and MLLMs remain limited by context length and weak instruction following, yet MLLMs are promising on shorter audio.

Audio inpainting aims to reconstruct missing segments in corrupted recordings. Most of existing methods produce plausible reconstructions when the gap lengths are short, but struggle to reconstruct gaps larger than about 100 ms. This paper explores recent advancements in deep learning and, particularly, diffusion models, for the task of audio inpainting. The proposed method uses an unconditionally trained generative model, which can be conditioned in a zero-shot fashion for audio inpainting, and is able to regenerate gaps of any size. An improved deep neural network architecture based on the constant-Q transform, which allows the model to exploit pitch-equivariant symmetries in audio, is also presented. The performance of the proposed algorithm is evaluated through objective and subjective metrics for the task of reconstructing short to mid-sized gaps, up to 300 ms. The results of a formal listening test show that the proposed method delivers comparable performance against the compared baselines for short gaps, such as 50 ms, while retaining a good audio quality and outperforming the baselines for wider gaps that are up to 300 ms long. The method presented in this paper can be applied to restoring sound recordings that suffer from severe local disturbances or dropouts, which must be reconstructed.

This paper presents BUT ReverbDB - a dataset of real room impulse responses (RIR), background noises and re-transmitted speech data. The retransmitted data includes LibriSpeech test-clean, 2000 HUB5 English evaluation and part of 2010 NIST Speaker Recognition Evaluation datasets. We provide a detailed description of RIR collection (hardware, software, post-processing) that can serve as a "cook-book" for similar efforts. We also validate BUT ReverbDB in two sets of automatic speech recognition (ASR) experiments and draw conclusions for augmenting ASR training data with real and artificially generated RIRs. We show that a limited number of real RIRs, carefully selected to match the target environment, provide results comparable to a large number of artificially generated RIRs, and that both sets can be combined to achieve the best ASR results. The dataset is distributed for free under a non-restrictive license and it currently contains data from 8 rooms, which is growing. The distribution package also contains a Kaldi-based recipe for augmenting publicly available AMI close-talk meeting data and test the results on an AMI single distant microphone set, allowing it to reproduce our experiments.

Word-level forced alignment estimates when each transcript word occurs in an audio recording. It underpins text-based media editing, subtitling, speech-data curation, and phonetic analysis. Existing evaluations understate the difficulty of forced alignment by relying on short, clean speech, perfect transcripts, and metrics that obscure consequential alignment errors. In contrast, real-world media and data-processing pipelines operate on long and diverse recordings. Additionally, forced aligners often operate on error-prone automatic speech recognition (ASR) output. We address these gaps with improved evaluation metrics, a scoring protocol for real ASR transcripts, and AlignBench, a benchmark spanning diverse speaker, acoustic, and text conditions. We further introduce FuseAlign, a transformer-based aligner trained on large-scale pseudo-labeled speech with online label correction. FuseAlign performs joint contextualization of audio and text for the localization of coarse words. The model then refines boundaries at millisecond resolution and detects missing transcript words in the audio without lexicon-based or Viterbi decoding. On AlignBench, FuseAlign substantially outperforms all baselines and remains robust under real ASR transcripts. Ablations show that convolutional upsampling and EMA-snapshot label correction matter more than model properties such as parameter count.

We introduce the REAL-TSE Challenge, an IEEE SLT 2026 satellite challenge on target speaker extraction~(TSE) from real conversational recordings. Given a multi-speaker mixture and one or more enrollment utterances from a target speaker, participating systems must recover only the target speech. Unlike simulated read-speech benchmarks, REAL-TSE evaluates Mandarin and English recordings that contain natural overlap, reverberation, noise, channel mismatch, and conversational dynamics. The challenge defines two complementary tracks: an Online track for low-latency streaming extraction and an Offline track for full-context processing. Systems are evaluated with Token Error Rate (TER), Speaker Similarity (SpkSim), DNSMOS, and target-speaker activity F1. This overview paper describes the task definition, datasets, baselines, evaluation protocol, submitted systems, condition-wise findings, and lessons for future real-world TSE benchmarks.

Public benchmarks are important measures of Automatic Speech Recognition (ASR) model capabilities. However, by nature of being public, there is risk of models being optimized for these benchmarks in ways that do not generalize well to real-world data. We present a methodology for quantifying benchmark optimization, focusing on cases where the audio underdetermines the reference transcript. We identify three families of behavioral probes that reveal models' capabilities of reproducing benchmark reference spans despite underdetermined audio: reference disagreement, masked-number recovery, and orthographic switching. We find that the highest-scoring open source models output verbatim reference transcript spans even when the relevant audio is contradictory, masked, or ambiguous. Using a variety of mechanistic probes, we show that models respond to narrow acoustic cues to override the faithful representation of the audio in favor of a benchmark-optimized policy. We show the benchmark-optimized behavior can be causally manipulated via low-rank linear steering or simply appending audio to the end of a segment in some cases. Overall, our results indicate that high-performing models exhibit benchmark-conditioned behaviors that can inflate benchmark performance without reflecting improved general-purpose transcription ability.

The MeViS-Audio track asks a system to segment the objects described by a spoken motion expression throughout a video and to return empty masks when the described target is absent. We present a simple staged solution. Qwen3-ASR first converts speech into text. Several video mask tracks are then produced with complementary grounding and segmentation models. Instead of trusting a single prediction, we select the track that has the highest average mask agreement with the other candidates. A small set of explicit direction, count, and plural rules corrects queries that require more than ordinary single-object tracking. Finally, a video-level classifier combines visual, audio-visual, and within-video query scores to decide whether any target is present. The submitted system obtains 0.5952 J &F, 0.7931 no-target accuracy, 0.9205 target accuracy, and a final score of 0.769589. The challenge organizers notified our team that this result ranked first in the track.

As more speech technologies rely on a supervised deep learning approach with clean speech as the ground truth, a methodology to onboard said speech at scale is needed. However, this approach needs to minimize the dependency on human listening and annotation, only requiring a human-in-the-loop when needed. In this paper, we address this issue by outlining Speech Enhancement-based Curation Pipeline (SECP) which serves as a framework to onboard clean speech. This clean speech can then train a speech enhancement model, which can further refine the original dataset and thus close the iterative loop. By running two iterative rounds, we observe that enhanced output used as ground truth does not degrade model performance according to Delta_{PESQ}, a metric used in this paper. We also show through comparative mean opinion score (CMOS) based subjective tests that the highest and lowest bound of refined data is perceptually better than the original data.

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.

Music recordings often suffer from audio quality issues such as excessive reverberation, distortion, clipping, tonal imbalances, and a narrowed stereo image, especially when created in non-professional settings without specialized equipment or expertise. These problems are typically corrected using separate specialized tools and manual adjustments. In this paper, we introduce SonicMaster, the first unified generative model for music restoration and mastering that addresses a broad spectrum of audio artifacts with text-based control. SonicMaster is conditioned on natural language instructions to apply targeted enhancements, or can operate in an automatic mode for general restoration. To train this model, we construct the SonicMaster dataset, a large dataset of paired degraded and high-quality tracks by simulating common degradation types with nineteen degradation functions belonging to five enhancements groups: equalization, dynamics, reverb, amplitude, and stereo. Our approach leverages a flow-matching generative training paradigm to learn an audio transformation that maps degraded inputs to their cleaned, mastered versions guided by text prompts. Objective audio quality metrics demonstrate that SonicMaster significantly improves sound quality across all artifact categories. Furthermore, subjective listening tests confirm that listeners prefer SonicMaster's enhanced outputs over the original degraded audio, highlighting the effectiveness of our unified approach.

Multi-speaker localization and tracking using microphone array recording is of importance in a wide range of applications. One of the challenges with multi-speaker tracking is to associate direction estimates with the correct speaker. Most existing association approaches rely on spatial or spectral information alone, leading to performance degradation when one of these information channels is partially known or missing. This paper studies a joint probability data association (JPDA)-based method that facilitates association based on joint spatial-spectral information. This is achieved by integrating speaker time-frequency (TF) masks, estimated based on spectral information, in the association probabilities calculation. An experimental study that tested the proposed method on recordings from the LOCATA challenge demonstrates the enhanced performance obtained by using joint spatial-spectral information in the association.

Generating lifelike conversational avatars requires modeling not just isolated speakers, but the dynamic, reciprocal interaction of speaking and listening. However, modeling the listener is exceptionally challenging: direct audio-driven training fails, producing stiff, static listening motions. This failure stems from a fundamental imbalance: the speaker's motion is strongly driven by speech audio, while the listener's motion primarily follows an internal motion prior and is only loosely guided by external speech. This challenge has led most methods to focus on speak-only generation. The only prior attempt at joint generation relies on extra speaker's motion to produce the listener. This design is not end-to-end, thereby hindering the real-time applicability. To address this limitation, we present UniLS, the first end-to-end framework for generating unified speak-listen expressions, driven by only dual-track audio. Our method introduces a novel two-stage training paradigm. Stage 1 first learns the internal motion prior by training an audio-free autoregressive generator, capturing the spontaneous dynamics of natural facial motion. Stage 2 then introduces the dual-track audio, fine-tuning the generator to modulate the learned motion prior based on external speech cues. Extensive evaluations show UniLS achieves state-of-the-art speaking accuracy. More importantly, it delivers up to 44.1\% improvement in listening metrics, generating significantly more diverse and natural listening expressions. This effectively mitigates the stiffness problem and provides a practical, high-fidelity audio-driven solution for interactive digital humans.

Masked token prediction has emerged as a powerful pre-training objective across language, vision, and speech, offering the potential to unify these diverse modalities through a single pre-training task. However, its application for general audio understanding remains underexplored, with BEATs being the only notable example. BEATs has seen limited modifications due to the absence of open-source pre-training code. Furthermore, BEATs was trained only on AudioSet, restricting its broader downstream applicability. To address these gaps, we present OpenBEATs, an open-source framework that extends BEATs via multi-domain audio pre-training. We conduct comprehensive evaluations across six types of tasks, twenty five datasets, and three audio domains, including audio reasoning tasks such as audio question answering, entailment, and captioning. OpenBEATs achieves state-of-the-art performance on six bioacoustics datasets, two environmental sound datasets and five reasoning datasets, performing better than models exceeding a billion parameters at one-fourth their parameter size. These results demonstrate the effectiveness of multi-domain datasets and masked token prediction task to learn general-purpose audio representations. To promote further research and reproducibility, we release all pre-training and evaluation code, pretrained and fine-tuned checkpoints, and training logs at https://shikhar-s.github.io/OpenBEATs

Recent language model (LM) advancements have showcased impressive zero-shot voice conversion (VC) performance. However, existing LM-based VC models usually apply offline conversion from source semantics to acoustic features, demanding the complete source speech, and limiting their deployment to real-time applications. In this paper, we introduce StreamVoice, a novel streaming LM-based model for zero-shot VC, facilitating real-time conversion given arbitrary speaker prompts and source speech. Specifically, to enable streaming capability, StreamVoice employs a fully causal context-aware LM with a temporal-independent acoustic predictor, while alternately processing semantic and acoustic features at each time step of autoregression which eliminates the dependence on complete source speech. To address the potential performance degradation from the incomplete context in streaming processing, we enhance the context-awareness of the LM through two strategies: 1) teacher-guided context foresight, using a teacher model to summarize the present and future semantic context during training to guide the model's forecasting for missing context; 2) semantic masking strategy, promoting acoustic prediction from preceding corrupted semantic and acoustic input, enhancing context-learning ability. Notably, StreamVoice is the first LM-based streaming zero-shot VC model without any future look-ahead. Experimental results demonstrate StreamVoice's streaming conversion capability while maintaining zero-shot performance comparable to non-streaming VC systems.

Diffusion and flow-matching models have revolutionized automatic text-to-audio generation in recent times. These models are increasingly capable of generating high quality and faithful audio outputs capturing to speech and acoustic events. However, there is still much room for improvement in creative audio generation that primarily involves music and songs. Recent open lyrics-to-song models, such as, DiffRhythm, ACE-Step, and LeVo, have set an acceptable standard in automatic song generation for recreational use. However, these models lack fine-grained word-level controllability often desired by musicians in their workflows. To the best of our knowledge, our flow-matching-based JAM is the first effort toward endowing word-level timing and duration control in song generation, allowing fine-grained vocal control. To enhance the quality of generated songs to better align with human preferences, we implement aesthetic alignment through Direct Preference Optimization, which iteratively refines the model using a synthetic dataset, eliminating the need or manual data annotations. Furthermore, we aim to standardize the evaluation of such lyrics-to-song models through our public evaluation dataset JAME. We show that JAM outperforms the existing models in terms of the music-specific attributes.

Speech translation models are increasingly capable of preserving speech-specific information (e.g., speaker gender, prosody, and emphasis), yet evaluation metrics remain blind to such phenomena. We meta-evaluate both text- and speech-based quality estimation metrics on two contrastive datasets targeting gender agreement and prosody, and find that both fall short, even when given direct access to the speech signal. We then train SpeechCOMET, a family of quality estimation models with speech encoders, and evaluate a state-of-the-art SpeechLLM as a judge. Both match or exceed text-based COMET on standard quality estimation, but neither consistently assesses speech-specific phenomena. We identify three causes: (1) speech-specific features are not reliably preserved in current encoders, (2) models tend to ignore the speech source signal, and (3) quality estimation training data contains too few relevant examples. We release all models and code, and argue that progress requires dedicated speech-specific training data and models that genuinely condition on speech.

Audio intelligence involves understanding, reasoning about, and generating both audio and speech. In this work, we introduce Nemotron-Labs-Audex-30B-A3B (Audex), a unified audio-text LLM built on Nemotron-Cascade-2-30B-A3B, a strong text-only MoE LLM. Audex adopts a simple unified design with a single Transformer decoder: audio inputs are encoded and projected into the text embedding space, while text tokens and quantized audio output tokens are treated uniformly during generation. This architecture enables strong audio-text fusion, seamless multimodal generation, and compatibility with standard LLM training and inference infrastructure. For training, we meticulously curate audio-text datasets comprising 157.4B audio tokens and 320.5B text tokens. We apply multi-stage supervised training on these datasets, followed by text-only Cascade RL and multi-domain on-policy distillation. Audex delivers state-of-the-art audio understanding, speech recognition and translation, text-to-speech, audio generation, and speech-to-speech generation, while preserving very compelling reasoning, alignment, knowledge, long-context, and agentic capabilities of its text-only LLM backbone with marginal or no regression. We release the model checkpoints to facilitate open research.

Understanding and explaining differences between audio recordings is crucial for fields like audio forensics, quality assessment, and audio generation. This involves identifying and describing audio events, acoustic scenes, signal characteristics, and their emotional impact on listeners. This paper stands out as the first work to comprehensively study the task of explaining audio differences and then propose benchmark, baselines for the task. First, we present two new datasets for audio difference explanation derived from the AudioCaps and Clotho audio captioning datasets. Using Large Language Models (LLMs), we generate three levels of difference explanations: (1) concise descriptions of audio events and objects, (2) brief sentences about audio events, acoustic scenes, and signal properties, and (3) comprehensive explanations that include semantics and listener emotions. For the baseline, we use prefix tuning where audio embeddings from two audio files are used to prompt a frozen language model. Our empirical analysis and ablation studies reveal that the naive baseline struggles to distinguish perceptually similar sounds and generate detailed tier 3 explanations. To address these limitations, we propose ADIFF, which introduces a cross-projection module, position captioning, and a three-step training process to enhance the model's ability to produce detailed explanations. We evaluate our model using objective metrics and human evaluation and show our model enhancements lead to significant improvements in performance over naive baseline and SoTA Audio-Language Model (ALM) Qwen Audio. Lastly, we conduct multiple ablation studies to study the effects of cross-projection, language model parameters, position captioning, third stage fine-tuning, and present our findings. Our benchmarks, findings, and strong baseline pave the way for nuanced and human-like explanations of audio differences.

The integration of audio perception capabilities into Large Language Models (LLMs) has enabled significant advances in Audio-LLMs. Although application-focused developments, particularly in curating training data for specific capabilities e.g., audio reasoning, have progressed rapidly, the underlying mechanisms that govern efficient transfer of rich semantic representations from audio encoders to LLMs remain under-explored. We conceptualize effective audio-LLM interaction as the LLM's ability to proficiently probe the audio encoder representations to satisfy textual queries. This paper presents a systematic investigation on how architectural design choices can affect that. Beginning with a standard Pengi/LLaVA-style audio-LLM architecture, we propose and evaluate several modifications guided by hypotheses derived from mechanistic interpretability studies and LLM operational principles. Our experiments demonstrate that: (1) delaying audio integration until the LLM's initial layers establish textual context that enhances its ability to probe the audio representations for relevant information; (2) the LLM can proficiently probe audio representations exclusively through LLM layer's attention submodule, without requiring propagation to its Feed-Forward Network (FFN) submodule; (3) an efficiently integrated ensemble of diverse audio encoders provides richer, complementary representations, thereby broadening the LLM's capacity to probe a wider spectrum of audio information. All hypotheses are evaluated using an identical three-stage training curriculum on a dataset of 5.6 million audio-text pairs, ensuring controlled comparisons. Our final architecture, which incorporates all proposed modifications, achieves relative improvements from 10\% to 60\% over the baseline, validating our approach to optimizing cross-modal information transfer in audio-LLMs. Project page: https://ta012.github.io/PAL/

Text-to-audio grounding (TAG) task aims to predict the onsets and offsets of sound events described by natural language. This task can facilitate applications such as multimodal information retrieval. This paper focuses on weakly-supervised text-to-audio grounding (WSTAG), where frame-level annotations of sound events are unavailable, and only the caption of a whole audio clip can be utilized for training. WSTAG is superior to strongly-supervised approaches in its scalability to large audio-text datasets. Two WSTAG frameworks are studied in this paper: sentence-level and phrase-level. First, we analyze the limitations of mean pooling used in the previous WSTAG approach and investigate the effects of different pooling strategies. We then propose phrase-level WSTAG to use matching labels between audio clips and phrases for training. Advanced negative sampling strategies and self-supervision are proposed to enhance the accuracy of the weak labels and provide pseudo strong labels. Experimental results show that our system significantly outperforms the previous WSTAG SOTA. Finally, we conduct extensive experiments to analyze the effects of several factors on phrase-level WSTAG. The code and model is available at https://github.com/wsntxxn/TextToAudioGrounding.

Generative modeling has recently achieved remarkable success across text, image, and audio domains, demonstrating powerful capabilities for unified representation learning. However, audio generation models still face challenges in terms of audio quality and generalization ability across tasks. This fragmentation results in redundant development efforts, inconsistent performance, and limited extensibility. To address these issues, we propose UniTok-Audio, a scalable and extensible framework for unified audio generation tasks. Specifically, 1) UniTok-Audio extracts continuous feature of conditions to generates discrete tokens of target audio in an autoregressive manner; 2) a special task identifier token unifies different learning patterns of multiple tasks in a single framework; 3) a dual-stream audio codec involving acoustic and semantic branch is developed for high-fidelity waveform reconstruction. Experimental results demonstrate that UniTok-Audio achieves competitive performance in comparation with state-of-the-art task-specific or multi-task systems across five time-aligned tasks: speech restoration, target speaker extraction, speech separation, voice conversion, and language-queried audio source separation. To foster future research, we will open-source our codebase. The demo page of our work can be found here: https://alibaba.github.io/unified-audio.

Speech Language Models (SLMs) are increasingly deployed in multi-speaker environments, yet their ability to attribute speech to the correct speaker and reason over speaker identities remains unclear. Hence, we introduce HEAR, a conceptually hierarchical benchmark diagnosing the foundational capabilities of speaker-attributed reasoning, comprising 2.4K human-verified samples from 887 diverse multi-party audio clips. Evaluating 20 leading SLMs on HEAR reveals they struggle with these foundational tasks, often relying on semantic priors rather than actual vocal cues. To address this, we present A2R, a 30B model optimized on Counterfactual Audio with Speaker-level Hard negatives (CASH), a dataset designed to guide the model to prioritize acoustic vocal cues over linguistic signals. A2R achieves strong performance on HEAR and exhibits zero-shot generalization to diverse multi-speaker downstream tasks, demonstrating that learned speaker attribution unlocks the model's latent capacity for speaker-aware reasoning. All resources are available at https://attributetoreason.github.io/AttributeToReason/

Audio-conditioned language models often underuse acoustic cues such as prosody, emotion, and non-speech sounds, raising the question of whether ASR-supervised frontends discard this information before it reaches the LM. We test whether the frontend is responsible by comparing Whisper-Tiny and Whisper-Small with EnCodec, DAC-VAE, and WavTokenizer in a shared Qwen3.5-4B audio-LM pipeline on ASR, emotion recognition, and sound captioning. Encoder replacement alone does not resolve this underuse: Whisper variants remain strongest overall, including on emotion and environmental sound captioning. To localize the failure, we trace task-relevant information through the encoder, projector, LM layers, and LM head. Linear probes and geometric analyses show that discriminative acoustic structure remains recoverable at the final LM layer, even when MCQA accuracy trails probe accuracy by up to 83 points. Because the answer format and decoding procedure are controlled, this task-dependent gap points to content-specific readout failure rather than generic format bias. LogitLens analyses and a targeted LM head intervention support the conclusion that acoustic underuse is not explained solely by encoder-side information loss and that readout alignment can be a dominant bottleneck.

Recent advances in Text-To-Speech (TTS) technology have enabled synthetic speech to mimic human voices with remarkable realism, raising significant security concerns. This underscores the need for traceable TTS models-systems capable of tracing their synthesized speech without compromising quality or security. However, existing methods predominantly rely on explicit watermarking on speech or on vocoder, which degrades speech quality and is vulnerable to spoofing. To address these limitations, we propose a novel framework for model attribution. Instead of embedding watermarks, we train the TTS model and discriminator using a joint training method that significantly improves traceability generalization while preserving-and even slightly improving-audio quality. This is the first work toward watermark-free TTS with strong traceability. To promote progress in related fields, we will release the code upon acceptance of the paper.

Sound designers and Foley artists usually sonorize a scene, such as from a movie or video game, by manually annotating and sonorizing each action of interest in the video. In our case, the intent is to leave full creative control to sound designers with a tool that allows them to bypass the more repetitive parts of their work, thus being able to focus on the creative aspects of sound production. We achieve this presenting Stable-V2A, a two-stage model consisting of: an RMS-Mapper that estimates an envelope representative of the audio characteristics associated with the input video; and Stable-Foley, a diffusion model based on Stable Audio Open that generates audio semantically and temporally aligned with the target video. Temporal alignment is guaranteed by the use of the envelope as a ControlNet input, while semantic alignment is achieved through the use of sound representations chosen by the designer as cross-attention conditioning of the diffusion process. We train and test our model on Greatest Hits, a dataset commonly used to evaluate V2A models. In addition, to test our model on a case study of interest, we introduce Walking The Maps, a dataset of videos extracted from video games depicting animated characters walking in different locations. Samples and code available on our demo page at https://ispamm.github.io/Stable-V2A.

Temporal grounding in long recordings remains challenging for audio-conditioned LLMs. We present a time-aware audio LLM that answers questions with explicit timestamps over up to 120 minutes of input. Our approach interleaves periodic time markers with continuous audio tokens using large-scale synthetic supervision from a cascaded pipeline. Our model achieves strong temporal-grounding accuracy on short and long benchmarks and supports time-anchored fragment descriptions and summaries. Extensive ablations examine how time representation, marker frequency, tokenization, and duration-mixture design affect accuracy and computational cost. We release model weights and datasets to support further research on time-aware audio understanding, available at https://huggingface.co/ai-sage/GigaChat3.1-Audio-10B-A1.8B.

Natural turn-taking in full-duplex voice interaction requires determining from partial speech whether a pause reflects hesitation or a completed conversational intent. Acoustic voice activity detection lacks this semantic information, while cascaded ASR-based endpointing introduces transcription dependence and additional processing stages. We formulate semantic endpoint detection as a causal audio-language reasoning task and introduce FD-VAD, an ASR-free streaming endpointer that maps bounded causal audio windows directly to Continue/Stop decisions. FD-VAD combines a frozen speech encoder with a lightweight modality adapter and a parameter-efficiently adapted language model, using a last-chunk training objective for streaming inference. We further introduce confidence-gated endpoint commitment to control interruption versus delay and boundary-focused hard-negative sampling to improve decisions around ambiguous turn boundaries. Across in-domain and conversational evaluations, FD-VAD outperforms strong streaming and non-streaming semantic turn classifiers, and achieves the highest EOT recall among qualifying systems on TurnBench dev set 0.853 (at FP<=0.10) in a zero-shot setting. These results show that semantic endpointing can be performed directly from streaming audio without intermediate ASR or dialogue state tracking.

In daily life, people hear speech, footsteps, and music around them. We can often recognize these sounds and judge where they come from. Each sound source can be shown on a separate acoustic map, a rectangular image covering 360^{circ} horizontally and 180^{circ} vertically. The map shows the directions occupied by the source as a region and the sound energy within that region. A class label identifies the sound. Predicting these labeled acoustic maps from audio is called semantic acoustic imaging. Such maps could help robots perceive their surroundings and allow augmented reality displays to show sound regions and classes over the real world. Existing models can recognize sound classes and estimate a direction for each source. However, a direction alone does not describe the source region or its energy. Acoustic imaging must also distinguish sound sources in nearby directions, while the number of active sources and the regions they occupy can change over time. We therefore propose the Semantic Acoustic Imaging Detector (SAID), which predicts a separate labeled acoustic map for each active source from audio. First, we pretrain Audio2Sph, SAID's audio encoder, through sound energy estimation across directions without class labels. Then, we train the complete SAID model to predict source regions, energy, and classes together. We also develop a pipeline that generates simulated recordings for pretraining and supports fine-tuning on real recordings. On the official DCASE2026 Task 3 Track A evaluation set, our submitted system ranks first with 0.1080 macro-averaged mean average precision (Macro mAP) and 0.3962 Macro Pearson r. Demos and code are provided at https://github.com/IN03X/SAID.

The scope of the International Workshop on Spoken Language Translation (IWSLT) has recently broadened beyond traditional Speech Translation (ST) to encompass a wider array of tasks, including Speech Question Answering and Summarization. This shift is partly driven by the growing capabilities of modern systems, particularly with the success of Large Language Models (LLMs). In this paper, we present the Karlsruhe Institute of Technology's submissions for the Offline ST and Instruction Following (IF) tracks, where we leverage LLMs to enhance performance across all tasks. For the Offline ST track, we propose a pipeline that employs multiple automatic speech recognition systems, whose outputs are fused using an LLM with document-level context. This is followed by a two-step translation process, incorporating additional refinement step to improve translation quality. For the IF track, we develop an end-to-end model that integrates a speech encoder with an LLM to perform a wide range of instruction-following tasks. We complement it with a final document-level refinement stage to further enhance output quality by using contextual information.

Recently, there has been an increasing focus on audio-text cross-modal learning. However, most of the existing audio-text datasets contain only simple descriptions of sound events. Compared with classification labels, the advantages of such descriptions are significantly limited. In this paper, we first analyze the detailed information that human descriptions of audio may contain beyond sound event labels. Based on the analysis, we propose an automatic pipeline for curating audio-text pairs with rich details. Leveraging the property that sounds can be mixed and concatenated in the time domain, we control details in four aspects: temporal relationship, loudness, speaker identity, and occurrence number, in simulating audio mixtures. Corresponding details are transformed into captions by large language models. Audio-text pairs with rich details in text descriptions are thereby obtained. We validate the effectiveness of our pipeline with a small amount of simulated data, demonstrating that the simulated data enables models to learn detailed audio captioning.

We present DPDFNet, a causal single-channel speech enhancement model that extends DeepFilterNet2 architecture with dual-path blocks in the encoder, strengthening long-range temporal and cross-band modeling while preserving the original enhancement framework. In addition, we demonstrate that adding a loss component to mitigate over-attenuation in the enhanced speech, combined with a fine-tuning phase tailored for "always-on" applications, leads to substantial improvements in overall model performance. To compare our proposed architecture with a variety of causal open-source models, we created a new evaluation set comprising long, low-SNR recordings in 12 languages across everyday noise scenarios, better reflecting real-world conditions than commonly used benchmarks. On this evaluation set, DPDFNet delivers superior performance to other causal open-source models, including some that are substantially larger and more computationally demanding. We also propose an holistic metric named PRISM, a composite, scale-normalized aggregate of intrusive and non-intrusive metrics, which demonstrates clear scalability with the number of dual-path blocks. We further demonstrate on-device feasibility by deploying DPDFNet on Ceva-NeuPro-Nano edge NPUs. Results indicate that DPDFNet-4, our second-largest model, achieves real-time performance on NPN32 and runs even faster on NPN64, confirming that state-of-the-art quality can be sustained within strict embedded power and latency constraints.

REPresentation Alignment (REPA) improves the training of generative flow models by aligning intermediate hidden states with pretrained teacher features, but its effectiveness in token-conditioned audio Flow Matching critically depends on the choice of supervised layers, which is typically made heuristically based on the depth. In this work, we introduce Attribution-Guided REPresentation Alignment (AG-REPA), a novel causal layer selection strategy for representation alignment in audio Flow Matching. Firstly, we find that layers that best store semantic/acoustic information (high teacher-space similarity) are not necessarily the layers that contribute most to the velocity field that drives generation, and we call it Store-Contribute Dissociation (SCD). To turn this insight into an actionable training guidance, we propose a forward-only gate ablation (FoG-A) that quantifies each layer's causal contribution via the induced change in the predicted velocity field, enabling sparse layer selection and adaptive weighting for alignment. Across unified speech and general-audio training (LibriSpeech + AudioSet) under different token-conditioning topologies, AG-REPA consistently outperforms REPA baselines. Overall, our results show that alignment is most effective when applied to the causally dominant layers that drive the velocity field, rather than to layers that are representationally rich but functionally passive.

We present LongCat-AudioDiT, a novel, non-autoregressive diffusion-based text-to-speech (TTS) model that achieves state-of-the-art (SOTA) performance. Unlike previous methods that rely on intermediate acoustic representations such as mel-spectrograms, the core innovation of LongCat-AudioDiT lies in operating directly within the waveform latent space. This approach effectively mitigates compounding errors and drastically simplifies the TTS pipeline, requiring only a waveform variational autoencoder (Wav-VAE) and a diffusion backbone. Furthermore, we introduce two critical improvements to the inference process: first, we identify and rectify a long-standing training-inference mismatch; second, we replace traditional classifier-free guidance with adaptive projection guidance to elevate generation quality. Experimental results demonstrate that, despite the absence of complex multi-stage training pipelines or high-quality human-annotated datasets, LongCat-AudioDiT achieves SOTA zero-shot voice cloning performance on the Seed benchmark while maintaining competitive intelligibility. Specifically, our largest variant, LongCat-AudioDiT-3.5B, outperforms the previous SOTA model (Seed-TTS), improving the speaker similarity (SIM) scores from 0.809 to 0.818 on Seed-ZH, and from 0.776 to 0.797 on Seed-Hard. Finally, through comprehensive ablation studies and systematic analysis, we validate the effectiveness of our proposed modules. Notably, we investigate the interplay between the Wav-VAE and the TTS backbone, revealing the counterintuitive finding that superior reconstruction fidelity in the Wav-VAE does not necessarily lead to better overall TTS performance. Code and model weights are released to foster further research within the speech community.

The rise of "bedroom producers" has democratized music creation, while challenging producers to objectively evaluate their work. To address this, we present AI TrackMate, an LLM-based music chatbot designed to provide constructive feedback on music productions. By combining LLMs' inherent musical knowledge with direct audio track analysis, AI TrackMate offers production-specific insights, distinguishing it from text-only approaches. Our framework integrates a Music Analysis Module, an LLM-Readable Music Report, and Music Production-Oriented Feedback Instruction, creating a plug-and-play, training-free system compatible with various LLMs and adaptable to future advancements. We demonstrate AI TrackMate's capabilities through an interactive web interface and present findings from a pilot study with a music producer. By bridging AI capabilities with the needs of independent producers, AI TrackMate offers on-demand analytical feedback, potentially supporting the creative process and skill development in music production. This system addresses the growing demand for objective self-assessment tools in the evolving landscape of independent music production.

Recent advancements in audio tokenization have significantly enhanced the integration of audio capabilities into large language models (LLMs). However, audio understanding and generation are often treated as distinct tasks, hindering the development of truly unified audio-language models. While instruction tuning has demonstrated remarkable success in improving generalization and zero-shot learning across text and vision, its application to audio remains largely unexplored. A major obstacle is the lack of comprehensive datasets that unify audio understanding and generation. To address this, we introduce Audio-FLAN, a large-scale instruction-tuning dataset covering 80 diverse tasks across speech, music, and sound domains, with over 100 million instances. Audio-FLAN lays the foundation for unified audio-language models that can seamlessly handle both understanding (e.g., transcription, comprehension) and generation (e.g., speech, music, sound) tasks across a wide range of audio domains in a zero-shot manner. The Audio-FLAN dataset is available on HuggingFace and GitHub and will be continuously updated.

Listeners of long-form talk-audio content, such as podcast episodes, often find it challenging to understand the overall structure and locate relevant sections. A practical solution is to divide episodes into chapters--semantically coherent segments labeled with titles and timestamps. Since most episodes on our platform at Spotify currently lack creator-provided chapters, automating the creation of chapters is essential. Scaling the chapterization of podcast episodes presents unique challenges. First, episodes tend to be less structured than written texts, featuring spontaneous discussions with nuanced transitions. Second, the transcripts are usually lengthy, averaging about 16,000 tokens, which necessitates efficient processing that can preserve context. To address these challenges, we introduce PODTILE, a fine-tuned encoder-decoder transformer to segment conversational data. The model simultaneously generates chapter transitions and titles for the input transcript. To preserve context, each input text is augmented with global context, including the episode's title, description, and previous chapter titles. In our intrinsic evaluation, PODTILE achieved an 11% improvement in ROUGE score over the strongest baseline. Additionally, we provide insights into the practical benefits of auto-generated chapters for listeners navigating episode content. Our findings indicate that auto-generated chapters serve as a useful tool for engaging with less popular podcasts. Finally, we present empirical evidence that using chapter titles can enhance effectiveness of sparse retrieval in search tasks.

We propose VoiceTailor, a parameter-efficient speaker-adaptive text-to-speech (TTS) system, by equipping a pre-trained diffusion-based TTS model with a personalized adapter. VoiceTailor identifies pivotal modules that benefit from the adapter based on a weight change ratio analysis. We utilize Low-Rank Adaptation (LoRA) as a parameter-efficient adaptation method and incorporate the adapter into pivotal modules of the pre-trained diffusion decoder. To achieve powerful adaptation performance with few parameters, we explore various guidance techniques for speaker adaptation and investigate the best strategies to strengthen speaker information. VoiceTailor demonstrates comparable speaker adaptation performance to existing adaptive TTS models by fine-tuning only 0.25\% of the total parameters. VoiceTailor shows strong robustness when adapting to a wide range of real-world speakers, as shown in the demo.

Despite the strong performance of large audio language models (LALMs) in various tasks, exactly how and where they integrate acoustic features with textual context remains unclear. We adapt causal tracing to investigate the internal information flow of LALMs during audio comprehension. By conducting layer-wise and token-wise analyses across DeSTA, Qwen, and Voxtral, we evaluate the causal effects of individual hidden states. Layer-wise analysis identifies different fusion strategies, from progressive integration in DeSTA to abrupt late-stage fusion in Qwen. Token-wise analysis shows that the final sequence token acts as an informational bottleneck where the network decisively retrieves relevant information from the audio. We also observe an attention-like query mechanism at intermediate token positions that triggers the model to pull task-relevant audio context. These findings provide a clear characterization of when and where multi-modal integration occurs within LALMs.

Neural speech codecs are increasingly used as intermediate representations in codec-based speech generation systems. TiCodec introduces a factorized representation that separates time-varying speech content from time-invariant information through a Time-Invariant Representation Extraction (TIRE) module, potentially reducing the amount of information that must be modeled at the frame-level. In this work, we investigate the nature of the information captured by TIRE representations and their suitability for low-latency speech processing. Using a series of probing tasks, we analyze the influence of the encoder layer and show that intermediate layers capture complementary speaker- and environment-related information while containing little linguistic content. We further study several segment selection strategies for TIRE training and demonstrate that cross-file sampling improves the robustness of invariant representations. Based on these findings, we propose Dual-TIRE, a multi-level architecture that exploits the complementarity of different encoder layers and improves speech reconstruction quality and speaker similarity. Finally, we evaluate TiCodec in a streaming inference setting using successive 660ms processing blocks. Results show that streaming operation can be achieved without significant degradation in reconstruction performance, highlighting the potential of factorized neural codec representations for future low-latency speech generation systems.

Recent advancements in song generation have shown promising results in generating songs from lyrics and/or global text prompts. However, most existing systems lack the ability to model the temporally varying attributes of songs, limiting fine-grained control over musical structure and dynamics. In this paper, we propose SegTune, a non-autoregressive framework for structured and controllable song generation. SegTune enables segment-level control by allowing users or large language models to specify local musical descriptions aligned to song sections.The segmental prompts are injected into the model by temporally broadcasting them to corresponding time windows, while global prompts influence the whole song to ensure stylistic coherence. To obtain accurate segment durations and enable precise lyric-to-music alignment, we introduce an LLM-based duration predictor that autoregressively generates sentence-level timestamped lyrics in LRC format. We further construct a large-scale data pipeline for collecting high-quality songs with aligned lyrics and prompts, and propose new evaluation metrics to assess segment-level alignment and vocal attribute consistency. Experimental results show that SegTune achieves superior controllability and musical coherence compared to existing baselines. See https://cai525.github.io/SegTune_demo for demos of our work.

The URGENT 2024 Challenge aims to foster speech enhancement (SE) techniques with great universality, robustness, and generalizability, featuring a broader task definition, large-scale multi-domain data, and comprehensive evaluation metrics. Nourished by the challenge outcomes, this paper presents an in-depth analysis of two key, yet understudied, issues in SE system development: data cleaning and evaluation metrics. We highlight several overlooked problems in traditional SE pipelines: (1) mismatches between declared and effective audio bandwidths, along with label noise even in various "high-quality" speech corpora; (2) lack of both effective SE systems to conquer the hardest conditions (e.g., speech overlap, strong noise / reverberation) and reliable measure of speech sample difficulty; (3) importance of combining multifaceted metrics for a comprehensive evaluation correlating well with human judgment. We hope that this endeavor can inspire improved SE pipeline designs in the future.

Recent advancements in audio generation have enabled the creation of high-fidelity audio clips from free-form textual descriptions. However, temporal relationships, a critical feature for audio content, are currently underrepresented in mainstream models, resulting in an imprecise temporal controllability. Specifically, users cannot accurately control the timestamps of sound events using free-form text. We acknowledge that a significant factor is the absence of high-quality, temporally-aligned audio-text datasets, which are essential for training models with temporal control. The more temporally-aligned the annotations, the better the models can understand the precise relationship between audio outputs and temporal textual prompts. Therefore, we present a strongly aligned audio-text dataset, AudioTime. It provides text annotations rich in temporal information such as timestamps, duration, frequency, and ordering, covering almost all aspects of temporal control. Additionally, we offer a comprehensive test set and evaluation metric to assess the temporal control performance of various models. Examples are available on the https://zeyuxie29.github.io/AudioTime/

Text-to-audio (TTA) generation aims to synthesize realistic audio that faithfully reflects natural-language descriptions. Most TTA systems adopt a two-stage latent paradigm: an audio tokenizer is optimized for reconstruction and then frozen, after which a generative model is trained in the resulting latent space. However, reconstruction-oriented representations may be suboptimal for generation, motivating joint representation and generative learning. To this end, we introduce Unite-Audio, to our knowledge, is the first to jointly learn continuous audio representations and latent flow matching for TTA. By coupling reconstruction with self-supervised generative prediction, Unite-Audio allows the generative objective to directly shape the latent space rather than treating it as a fixed intermediate representation. We further employ Flow-GRPO post-training to improve text-conditioned generation. Experiments show competitive TTA performance with a compact latent flow model, while ablation studies confirm the benefit of jointly learning the audio representation and generative model. Audio samples are available at https://runwushi.github.io/Unite-Audio.

The training of deep learning-based multichannel speech enhancement and source localization systems relies heavily on the simulation of room impulse response and multichannel diffuse noise, due to the lack of large-scale real-recorded datasets. However, the acoustic mismatch between simulated and real-world data could degrade the model performance when applying in real-world scenarios. To bridge this simulation-to-real gap, this paper presents a new relatively large-scale Real-recorded and annotated Microphone Array speech&Noise (RealMAN) dataset. The proposed dataset is valuable in two aspects: 1) benchmarking speech enhancement and localization algorithms in real scenarios; 2) offering a substantial amount of real-world training data for potentially improving the performance of real-world applications. Specifically, a 32-channel array with high-fidelity microphones is used for recording. A loudspeaker is used for playing source speech signals. A total of 83-hour speech signals (48 hours for static speaker and 35 hours for moving speaker) are recorded in 32 different scenes, and 144 hours of background noise are recorded in 31 different scenes. Both speech and noise recording scenes cover various common indoor, outdoor, semi-outdoor and transportation environments, which enables the training of general-purpose speech enhancement and source localization networks. To obtain the task-specific annotations, the azimuth angle of the loudspeaker is annotated with an omni-direction fisheye camera by automatically detecting the loudspeaker. The direct-path signal is set as the target clean speech for speech enhancement, which is obtained by filtering the source speech signal with an estimated direct-path propagation filter.

We introduce a new collection of spoken English audio suitable for training speech recognition systems under limited or no supervision. It is derived from open-source audio books from the LibriVox project. It contains over 60K hours of audio, which is, to our knowledge, the largest freely-available corpus of speech. The audio has been segmented using voice activity detection and is tagged with SNR, speaker ID and genre descriptions. Additionally, we provide baseline systems and evaluation metrics working under three settings: (1) the zero resource/unsupervised setting (ABX), (2) the semi-supervised setting (PER, CER) and (3) the distant supervision setting (WER). Settings (2) and (3) use limited textual resources (10 minutes to 10 hours) aligned with the speech. Setting (3) uses large amounts of unaligned text. They are evaluated on the standard LibriSpeech dev and test sets for comparison with the supervised state-of-the-art.

Full-duplex dialogue requires timely yet selective interruption handling, which end-of-turn prediction alone cannot achieve: complete utterances may need no response, while unfinished requests may warrant interruption. To address this challenge, we propose HiThink Turn, an intent-aware streaming turn-state predictor that separates response intent from semantic completeness and conditions decisions on system playback state. A key contribution is minimal intent-sufficient prefix supervision, constructed through LLM judgments and speech alignment, while training on audio truncated at chunk boundaries improves robustness to partial speech. These components support streaming inference with 240-ms audio chunks, enabling low-latency, accurate full-duplex turn control. Experiments show that HiThink Turn leads the compared methods in Easy Turn macro accuracy, Full-Duplex-Bench average interaction rate score (0.933), and non-target-speech average playback resume rate (0.735). Additionally, intent-prefix triggering raises interruption success from 89\% to 98\% and reduces mean stop latency by 60.9\%.

While speech quality is typically assessed on complete utterances, streaming and generative systems require incremental estimation from partial audio. Existing predictors assume full context, degrading on prefix-constrained inputs. Extending ARECHO, we propose ANCHOR, reformulating incremental assessment as a multi-resolution autoregressive task. It models chunk- and utterance-level quality within a single decoder using dual-resolution tokens and a resolution-aware hierarchy for coarse-to-fine refinement. Experiments show substantial robustness under partial input, including a 48% PLCMOS error reduction on 2-second prefixes. Convergence analysis reveals a 4-6 s effective perceptual context horizon. A stress test further isolates structured extrapolation biases under localized corruption. Results demonstrate that hierarchical supervision improves incremental prediction and elucidates how perceptual quality accumulates over time.

Processing long-form audio is a major challenge for Large Audio Language models (LALMs). These models struggle with the quadratic cost of attention (O(N^2)) and with modeling long-range temporal dependencies. Existing audio benchmarks are built mostly from short clips and do not evaluate models in realistic long context settings. To address this gap, we introduce AudioMarathon, a benchmark designed to evaluate both understanding and inference efficiency on long-form audio. AudioMarathon provides a diverse set of tasks built upon three pillars: long-context audio inputs with durations ranging from 90.0 to 300.0 seconds, which correspond to encoded sequences of 2,250 to 7,500 audio tokens, respectively, full domain coverage across speech, sound, and music, and complex reasoning that requires multi-hop inference. We evaluate state-of-the-art LALMs and observe clear performance drops as audio length grows. We also study acceleration techniques and analyze the trade-offs of token pruning and KV cache eviction. The results show large gaps across current LALMs and highlight the need for better temporal reasoning and memory-efficient architectures. We believe AudioMarathon will drive the audio and multimodal research community to develop more advanced audio understanding models capable of solving complex audio tasks.

Video dubbing is a cornerstone of multimedia content creation, aiming to synthesize synchronized acoustic sequences for visual streams. While Text-to-Speech (TTS) and Text-to-Audio (TTA) generation have each achieved remarkable progress, existing dubbing systems remain confined to isolated speech synthesis without incorporating sound effects and ambient audio, forcing practitioners to rely on fragmented workflows and laborious manual post-mixing. To address this limitation, we present HoliDubber, a holistic video dubbing framework that moves beyond speech-only generation by enabling the joint synthesis of speech and sound effects from a single text prompt. Specifically, HoliDubber adopts a patch-based autoregressive diffusion transformer architecture, where a causal language model autoregressively models aggregated patch embeddings to capture global temporal structure, and a Diffusion Transformer decoder generates high-fidelity continuous tokens within each patch, following a divide-and-conquer strategy. To achieve cross-modal alignment, visual features are encoded into patch-level representations and fused with audio patches via cross-attention, enabling the model to ground speech generation in the speaker's visual articulation dynamics. In addition, we introduce HoliDub-Bench, a benchmark curated from established datasets with synchronized video-text-audio triplets designed for holistic dubbing evaluation. Extensive experiments demonstrate that HoliDubber significantly outperforms existing methods across multiple benchmarks in speech quality, synchronization, and speaker similarity. Furthermore, results on HoliDub-Bench validate the effectiveness of joint speech-and-sound generation, establishing a new paradigm for holistic video dubbing in complex acoustic scenes. The demo page of the project is https://holidubber.github.io

This paper presents an innovative approach to enhance control over audio generation by emphasizing the alignment between audio and text representations during model training. In the context of language model-based audio generation, the model leverages input from both textual and audio token representations to predict subsequent audio tokens. However, the current configuration lacks explicit regularization to ensure the alignment between the chosen text representation and the language model's predictions. Our proposal involves the incorporation of audio and text representation regularization, particularly during the classifier-free guidance (CFG) phase, where the text condition is excluded from cross attention during language model training. The aim of this proposed representation regularization is to minimize discrepancies in audio and text similarity compared to other samples within the same training batch. Experimental results on both music and audio generation tasks demonstrate that our proposed methods lead to improvements in objective metrics for both audio and music generation, as well as an enhancement in the human perception for audio generation.

Large-scale pre-trained image-text models demonstrate remarkable versatility across diverse tasks, benefiting from their robust representational capabilities and effective multimodal alignment. We extend the application of these models, specifically CLIP, to the domain of sound source localization. Unlike conventional approaches, we employ the pre-trained CLIP model without explicit text input, relying solely on the audio-visual correspondence. To this end, we introduce a framework that translates audio signals into tokens compatible with CLIP's text encoder, yielding audio-driven embeddings. By directly using these embeddings, our method generates audio-grounded masks for the provided audio, extracts audio-grounded image features from the highlighted regions, and aligns them with the audio-driven embeddings using the audio-visual correspondence objective. Our findings suggest that utilizing pre-trained image-text models enable our model to generate more complete and compact localization maps for the sounding objects. Extensive experiments show that our method outperforms state-of-the-art approaches by a significant margin.

Spoken semantic parsing (SSP) involves generating machine-comprehensible parses from input speech. Training robust models for existing application domains represented in training data or extending to new domains requires corresponding triplets of speech-transcript-semantic parse data, which is expensive to obtain. In this paper, we address this challenge by examining methods that can use transcript-semantic parse data (unpaired text) without corresponding speech. First, when unpaired text is drawn from existing textual corpora, Joint Audio Text (JAT) and Text-to-Speech (TTS) are compared as ways to generate speech representations for unpaired text. Experiments on the STOP dataset show that unpaired text from existing and new domains improves performance by 2% and 30% in absolute Exact Match (EM) respectively. Second, we consider the setting when unpaired text is not available in existing textual corpora. We propose to prompt Large Language Models (LLMs) to generate unpaired text for existing and new domains. Experiments show that examples and words that co-occur with intents can be used to generate unpaired text with Llama 2.0. Using the generated text with JAT and TTS for spoken semantic parsing improves EM on STOP by 1.4% and 2.6% absolute for existing and new domains respectively.

Recent audio-language models are increasingly evaluated on recognizing, describing, and reasoning about audio, but expert-facing applications require a different capability: producing feedback that identifies problems and suggests corrective actions grounded in the input. We introduce VocalCoachBench, a benchmark for evaluating audio-language models on expert vocal coaching feedback for singing. VocalCoachBench contains 515 recordings annotated by 18 professional vocal trainers, yielding 1,056 expert submissions and 12,051 atomic coaching claims. It comprises a same-song subset for controlled comparison and a diverse-song subset for segment-grounded feedback across varied songs and recording conditions. To accommodate the open-ended nature of expert feedback, VocalCoachBench sep- arates deterministic structured targets from claim-based assessment of free-form diagnosis and corrective guidance. Human annotation analysis shows that expert agreement varies strongly with label granularity, motivating hierarchical structured metrics and claim-based evaluation of open-ended feedback. Experiments with 12 recent audio-language models reveal a consistent gap: while models can compare performances and identify broad issue domains in free-form feedback, Top-3 fine-grained issue-label identification remains below label-prior baselines and strict diagnosis alignment stays below 7%. To our knowledge, VocalCoachBench pro- vides the first public testbed for evaluating audio-grounded expert feedback for singing, moving audio-language evaluation beyond description toward analytic feedback.

In recent years, Text-to-Audio Generation has achieved remarkable progress, offering sound creators powerful tools to transform textual inspirations into vivid audio. However, existing models predominantly operate directly in the acoustic latent space of a Variational Autoencoder (VAE), often leading to suboptimal alignment between generated audio and textual descriptions. In this paper, we introduce SemanticAudio, a novel framework that conducts both audio generation and editing directly in a high-level semantic space. We define this semantic space as a compact representation capturing the global identity and temporal sequence of sound events, distinct from fine-grained acoustic details. SemanticAudio employs a two-stage Flow Matching architecture: the Semantic Planner first generates these compact semantic features to sketch the global semantic layout, and the Acoustic Synthesizer subsequently produces high-fidelity acoustic latents conditioned on this semantic plan. Leveraging this decoupled design, we further introduce a training-free text-guided editing mechanism that enables precise attribute-level modifications on general audio without retraining. Specifically, this is achieved by steering the semantic generation trajectory via the difference of velocity fields derived from source and target text prompts. Extensive experiments demonstrate that SemanticAudio surpasses existing mainstream approaches in semantic alignment. Demo available at: https://semanticaudio1.github.io/

CHILDES is a large-scale child speech corpus containing long-form recordings of naturalistic child-adult interactions, making it a valuable resource for studying child speech and language development. However, utterance-level timestamps provided in this corpus are often noisy, incomplete, or misaligned with the audio. As a result, utterances cannot always be reliably localized within long recordings, which limits the direct use of these data for training and evaluating speech models. In this work, we propose BEACON (Boundary Estimation via Alignment CONsensus), an ensemble timestamp-curation framework that refines utterance-level timestamps by aggregating knowledge from multiple off-the-shelf ASR models. Specifically, each model's word-level timestamp predictions are first aligned to provided human transcripts, and the final utterance time boundaries are determined by a consensus voting strategy. The framework is corpus-agnostic and applies to any long-form recording paired with a trusted transcript whose timestamps are unreliable or missing, offering a general recipe for timestamp curation. Leveraging this pipeline, we curate and release a 413-hour general-purpose child-speech dataset with corrected utterance-level timestamps, together with a 283-hour quality-controlled subset for ASR training. Fine-tuning on this subset yields up to an average 19.5% relative WER reduction on four out-of-domain child-speech benchmarks.

Audio inpainting refers to the task of reconstructing missing segments in corrupted audio recordings. While prior approaches-including waveform and spectrogram-based diffusion models-have shown promising results for short gaps, they often degrade in quality when gaps exceed 100 milliseconds (ms). In this work, we introduce a novel inpainting method based on discrete diffusion modeling, which operates over tokenized audio representations produced by a pre-trained audio tokenizer. Our approach models the generative process directly in the discrete latent space, enabling stable and semantically coherent reconstruction of missing audio. We evaluate the method on the MusicNet dataset using both objective and perceptual metrics across gap durations up to 300 ms. We further evaluated our approach on the MTG dataset, extending the gap duration to 500 ms. Experimental results demonstrate that our method achieves competitive or superior performance compared to existing baselines, particularly for longer gaps, offering a robust solution for restoring degraded musical recordings. Audio examples of our proposed method can be found at https://iftach21.github.io/

Interpretability of deep neural networks is a recently emerging area of machine learning research targeting a better understanding of how models perform feature selection and derive their classification decisions. This paper explores the interpretability of neural networks in the audio domain by using the previously proposed technique of layer-wise relevance propagation (LRP). We present a novel audio dataset of English spoken digits which we use for classification tasks on spoken digits and speaker's gender. We use LRP to identify relevant features for two neural network architectures that process either waveform or spectrogram representations of the data. Based on the relevance scores obtained from LRP, hypotheses about the neural networks' feature selection are derived and subsequently tested through systematic manipulations of the input data. The results confirm that the networks are highly reliant on features marked as relevant by LRP.

Speech and audio LLMs are evaluated by comparing waveform predictions with predictions from an automatic speech recognition (ASR) transcript. For fixed closed-set tasks, this conflates acoustic evidence with the need to invoke a generative audio model. We estimate incremental call value with matched selectors sharing pre-call evidence. Each policy may retain the transcript label, use a local encoder, or invoke a generative model; matched control removes generative actions but preserves pre-call evidence and development selection. On VocalSound, transcript-only accuracy is 0.296, while supervised CLAP and WavLM controls reach 0.850 and 0.854 without calls. Full selector reaches 0.925 at 12.5% calls versus 0.921 for matched No-call selector (difference 0.004; 95% CI [-0.025, 0.033]). Thus, results do not show a call gain after transcript and encoder evidence are available. Relevant quantity is incremental accuracy from allowing calls, not the waveform-transcript gap.

Neural text-to-speech (TTS) systems systematically mispronounce low-resource proper nouns, particularly non-English names, brands, and geographic locations, due to their underrepresentation in predominantly English training corpora. Existing solutions typically rely on expensive multilingual data collection, supervised finetuning, or manual phonetic annotation, which limits the deployment of TTS systems in linguistically diverse settings. We introduce SonoEdit, a model editing technique that surgically corrects pronunciation errors in pre-trained TTS models without retraining. Instead of costly finetuning or explicit phoneme injection, we propose a parsimonious alternative based on Null-Space Pronunciation Editing, which performs a single-shot parameter update to modify the pronunciation of specific words while provably preserving all other model behavior. We first adapt Acoustic Causal Tracing to identify the Transformer layers responsible for text-to-pronunciation mapping. We then apply Null-Space Constrained Editing to compute a closed-form weight update that corrects the target pronunciation while remaining mathematically orthogonal to the subspace governing general speech generation. This constrained update steers the model's acoustic output toward a desired pronunciation exemplar while guaranteeing zero first-order change on a preserved speech corpus.

Speech restoration (SR) is a task of converting degraded speech signals into high-quality ones. In this study, we propose a robust SR model called Miipher, and apply Miipher to a new SR application: increasing the amount of high-quality training data for speech generation by converting speech samples collected from the Web to studio-quality. To make our SR model robust against various degradation, we use (i) a speech representation extracted from w2v-BERT for the input feature, and (ii) a text representation extracted from transcripts via PnG-BERT as a linguistic conditioning feature. Experiments show that Miipher (i) is robust against various audio degradation and (ii) enable us to train a high-quality text-to-speech (TTS) model from restored speech samples collected from the Web. Audio samples are available at our demo page: google.github.io/df-conformer/miipher/

This paper presents a streaming extension for the Sortformer speaker diarization framework, whose key property is the arrival-time ordering of output speakers. The proposed approach employs an Arrival-Order Speaker Cache (AOSC) to store frame-level acoustic embeddings of previously observed speakers. Unlike conventional speaker-tracing buffers, AOSC orders embeddings by speaker index corresponding to their arrival time order, and is dynamically updated by selecting frames with the highest scores based on the model's past predictions. Notably, the number of stored embeddings per speaker is determined dynamically by the update mechanism, ensuring efficient cache utilization and precise speaker tracking. Experiments on benchmark datasets confirm the effectiveness and flexibility of our approach, even in low-latency setups. These results establish Streaming Sortformer as a robust solution for real-time multi-speaker tracking and a foundation for streaming multi-talker speech processing.

Zero-shot text-to-speech models can clone a speaker's timbre from a short reference audio, but they also strongly inherit the speaking style present in the reference. As a result, synthesizing speech with a desired style often requires carefully selecting reference audio, which is impractical when only limited or mismatched references are available. While recent controllable TTS methods attempt to address this issue, they typically rely on absolute style targets and discrete textual prompts, and therefore do not support continuous and reference-relative style control. We propose ReStyle-TTS, a framework that enables continuous and reference-relative style control in zero-shot TTS. Our key insight is that effective style control requires first reducing the model's implicit dependence on reference style before introducing explicit control mechanisms. To this end, we introduce Decoupled Classifier-Free Guidance (DCFG), which independently controls text and reference guidance, reducing reliance on reference style while preserving text fidelity. On top of this, we apply style-specific LoRAs together with Orthogonal LoRA Fusion to enable continuous and disentangled multi-attribute control, and introduce a Timbre Consistency Optimization module to mitigate timbre drift caused by weakened reference guidance. Experiments show that ReStyle-TTS enables user-friendly, continuous, and relative control over pitch, energy, and multiple emotions while maintaining intelligibility and speaker timbre, and performs robustly in challenging mismatched reference-target style scenarios.

Training data cleaning is a new application for generative model-based speech restoration (SR). This paper introduces Miipher-2, an SR model designed for million-hour scale data, for training data cleaning for large-scale generative models like large language models. Key challenges addressed include generalization to unseen languages, operation without explicit conditioning (e.g., text, speaker ID), and computational efficiency. Miipher-2 utilizes a frozen, pre-trained Universal Speech Model (USM), supporting over 300 languages, as a robust, conditioning-free feature extractor. To optimize efficiency and minimize memory, Miipher-2 incorporates parallel adapters for predicting clean USM features from noisy inputs and employs the WaveFit neural vocoder for waveform synthesis. These components were trained on 3,000 hours of multi-lingual, studio-quality recordings with augmented degradations, while USM parameters remained fixed. Experimental results demonstrate Miipher-2's superior or comparable performance to conventional SR models in word-error-rate, speaker similarity, and both objective and subjective sound quality scores across all tested languages. Miipher-2 operates efficiently on consumer-grade accelerators, achieving a real-time factor of 0.0078, enabling the processing of a million-hour speech dataset in approximately three days using only 100 such accelerators.

Large diffusion models have been successful in text-to-audio (T2A) synthesis tasks, but they often suffer from common issues such as semantic misalignment and poor temporal consistency due to limited natural language understanding and data scarcity. Additionally, 2D spatial structures widely used in T2A works lead to unsatisfactory audio quality when generating variable-length audio samples since they do not adequately prioritize temporal information. To address these challenges, we propose Make-an-Audio 2, a latent diffusion-based T2A method that builds on the success of Make-an-Audio. Our approach includes several techniques to improve semantic alignment and temporal consistency: Firstly, we use pre-trained large language models (LLMs) to parse the text into structured <event & order> pairs for better temporal information capture. We also introduce another structured-text encoder to aid in learning semantic alignment during the diffusion denoising process. To improve the performance of variable length generation and enhance the temporal information extraction, we design a feed-forward Transformer-based diffusion denoiser. Finally, we use LLMs to augment and transform a large amount of audio-label data into audio-text datasets to alleviate the problem of scarcity of temporal data. Extensive experiments show that our method outperforms baseline models in both objective and subjective metrics, and achieves significant gains in temporal information understanding, semantic consistency, and sound quality.

We propose a novel step-by-step video-to-audio generation method that sequentially produces individual audio tracks, each corresponding to a specific sound event in the video. Our approach mirrors traditional Foley workflows, aiming to capture all sound events induced by a given video comprehensively. Each generation step is formulated as a guided video-to-audio synthesis task, conditioned on a target text prompt and previously generated audio tracks. This design is inspired by the idea of concept negation from prior compositional generation frameworks. To enable this guided generation, we introduce a training framework that leverages pre-trained video-to-audio models and eliminates the need for specialized paired datasets, allowing training on more accessible data. Experimental results demonstrate that our method generates multiple semantically distinct audio tracks for a single input video, leading to higher-quality composite audio synthesis than existing baselines.

Query-based universal sound separation is fundamental to intelligent auditory systems, aiming to isolate specific sources from mixtures. Despite recent advances, existing methods continue to suffer from residual interference in complex acoustic scenes. This performance limitation stems largely from a data bottleneck: in-the-wild datasets contain weak labels and severe co-occurrence of events. These flaws induce models to learn spurious correlations between background noise and target categories instead of robust acoustic features. To address this, we propose an automated pipeline that eliminates co-occurrence of events by mining high-purity single-event segments from in-the-wild datasets via a semantically consistent synthesis protocol. Utilizing this pipeline, we constructed Hive, a high-quality synthetic dataset comprising 2.4k hours of raw audio. Experimental results demonstrate that, compared with the state-of-the-art model SAM-Audio which was trained on a huge dataset sim500 times larger than Hive, certain open-source models trained on Hive achieve competitive separation accuracy and perceptual quality. Moreover, these models exhibited remarkable zero-shot generalization on out-of-distribution evaluation benchmarks. These findings highlight that prioritizing purity of supervised signals enables significant data efficiency, offering a new paradigm for training robust auditory foundation models with reduced computational costs. Code and dataset are available at https://shandaai.github.io/Hive.

To reveal the importance of temporal precision in ground truth audio event labels, we collected precise (~0.1 sec resolution) "strong" labels for a portion of the AudioSet dataset. We devised a temporally strong evaluation set (including explicit negatives of varying difficulty) and a small strong-labeled training subset of 67k clips (compared to the original dataset's 1.8M clips labeled at 10 sec resolution). We show that fine-tuning with a mix of weak and strongly labeled data can substantially improve classifier performance, even when evaluated using only the original weak labels. For a ResNet50 architecture, d' on the strong evaluation data including explicit negatives improves from 1.13 to 1.41. The new labels are available as an update to AudioSet.

Forced alignment estimates the timestamps of each word, phone or character in speech given its transcript. Published comparisons normalize transcripts, split the data and match boundaries differently, so their numbers cannot be read together. We present FA-Bench, an open framework that fixes those choices once and releases the code, splits, phone mapping, text normalization and scoring script, with results published periodically. Track 1 gives every aligner the reference transcript and Track 2 gives it a recognizer's output, on the same audio, clean and degraded four ways, with 21 open models and 9 commercial APIs under a unified protocol. We score every boundary of an utterance and check the two labels beside it, so a word the recognizer missed or invented is charged. Using a tolerance-based F1 as our primary metric eliminates the 9% to 14% score inflation that standard MAE causes on recognition-dependent systems in conversational speech. We then group boundaries by position and how many adjacent words were recognized correctly, which shows where a system lost the score. We discovered systematic bias in how current systems time words, with Whisper about 150 ms early and several commercial ASR APIs over 50 ms late. Code and results are at https://github.com/olewave/fa-bench

Recently, instruction-following audio-language models have received broad attention for human-audio interaction. However, the absence of benchmarks capable of evaluating audio-centric interaction capabilities has impeded advancements in this field. Previous models primarily focus on assessing different fundamental tasks, such as Automatic Speech Recognition (ASR), and lack an assessment of the open-ended generative capabilities centered around audio. Thus, it is challenging to track the progression in the Large Audio-Language Models (LALMs) domain and to provide guidance for future improvement. In this paper, we introduce AIR-Bench (Audio InstRuction Benchmark), the first benchmark designed to evaluate the ability of LALMs to understand various types of audio signals (including human speech, natural sounds, and music), and furthermore, to interact with humans in the textual format. AIR-Bench encompasses two dimensions: foundation and chat benchmarks. The former consists of 19 tasks with approximately 19k single-choice questions, intending to inspect the basic single-task ability of LALMs. The latter one contains 2k instances of open-ended question-and-answer data, directly assessing the comprehension of the model on complex audio and its capacity to follow instructions. Both benchmarks require the model to generate hypotheses directly. We design a unified framework that leverages advanced language models, such as GPT-4, to evaluate the scores of generated hypotheses given the meta-information of the audio. Experimental results demonstrate a high level of consistency between GPT-4-based evaluation and human evaluation. By revealing the limitations of existing LALMs through evaluation results, AIR-Bench can provide insights into the direction of future research.

Recent advances in Text-to-Speech (TTS) and Voice-Conversion (VC) using generative Artificial Intelligence (AI) technology have made it possible to generate high-quality and realistic human-like audio. This poses growing challenges in distinguishing AI-synthesized speech from the genuine human voice and could raise concerns about misuse for impersonation, fraud, spreading misinformation, and scams. However, existing detection methods for AI-synthesized audio have not kept pace and often fail to generalize across diverse datasets. In this paper, we introduce SONAR, a synthetic AI-Audio Detection Framework and Benchmark, aiming to provide a comprehensive evaluation for distinguishing cutting-edge AI-synthesized auditory content. SONAR includes a novel evaluation dataset sourced from 9 diverse audio synthesis platforms, including leading TTS providers and state-of-the-art TTS models. It is the first framework to uniformly benchmark AI-audio detection across both traditional and foundation model-based detection systems. Through extensive experiments, (1) we reveal the limitations of existing detection methods and demonstrate that foundation models exhibit stronger generalization capabilities, likely due to their model size and the scale and quality of pretraining data. (2) Speech foundation models demonstrate robust cross-lingual generalization capabilities, maintaining strong performance across diverse languages despite being fine-tuned solely on English speech data. This finding also suggests that the primary challenges in audio deepfake detection are more closely tied to the realism and quality of synthetic audio rather than language-specific characteristics. (3) We explore the effectiveness and efficiency of few-shot fine-tuning in improving generalization, highlighting its potential for tailored applications, such as personalized detection systems for specific entities or individuals.

Universal sound separation aims to extract clean audio tracks corresponding to distinct events from mixed audio, which is critical for artificial auditory perception. However, current methods heavily rely on artificially mixed audio for training, which limits their ability to generalize to naturally mixed audio collected in real-world environments. To overcome this limitation, we propose ClearSep, an innovative framework that employs a data engine to decompose complex naturally mixed audio into multiple independent tracks, thereby allowing effective sound separation in real-world scenarios. We introduce two remix-based evaluation metrics to quantitatively assess separation quality and use these metrics as thresholds to iteratively apply the data engine alongside model training, progressively optimizing separation performance. In addition, we propose a series of training strategies tailored to these separated independent tracks to make the best use of them. Extensive experiments demonstrate that ClearSep achieves state-of-the-art performance across multiple sound separation tasks, highlighting its potential for advancing sound separation in natural audio scenarios. For more examples and detailed results, please visit our demo page at https://clearsep.github.io.

We propose a simple but efficient method termed Guided Learning for weakly-labeled semi-supervised sound event detection (SED). There are two sub-targets implied in weakly-labeled SED: audio tagging and boundary detection. Instead of designing a single model by considering a trade-off between the two sub-targets, we design a teacher model aiming at audio tagging to guide a student model aiming at boundary detection to learn using the unlabeled data. The guidance is guaranteed by the audio tagging performance gap of the two models. In the meantime, the student model liberated from the trade-off is able to provide more excellent boundary detection results. We propose a principle to design such two models based on the relation between the temporal compression scale and the two sub-targets. We also propose an end-to-end semi-supervised learning process for these two models to enable their abilities to rise alternately. Experiments on the DCASE2018 Task4 dataset show that our approach achieves competitive performance.

This paper introduces FALL-E, a foley synthesis system and its training/inference strategies. The FALL-E model employs a cascaded approach comprising low-resolution spectrogram generation, spectrogram super-resolution, and a vocoder. We trained every sound-related model from scratch using our extensive datasets, and utilized a pre-trained language model. We conditioned the model with dataset-specific texts, enabling it to learn sound quality and recording environment based on text input. Moreover, we leveraged external language models to improve text descriptions of our datasets and performed prompt engineering for quality, coherence, and diversity. FALL-E was evaluated by an objective measure as well as listening tests in the DCASE 2023 challenge Task 7. The submission achieved the second place on average, while achieving the best score for diversity, second place for audio quality, and third place for class fitness.

Removing background noise from speech audio has been the subject of considerable effort, especially in recent years due to the rise of virtual communication and amateur recordings. Yet background noise is not the only unpleasant disturbance that can prevent intelligibility: reverb, clipping, codec artifacts, problematic equalization, limited bandwidth, or inconsistent loudness are equally disturbing and ubiquitous. In this work, we propose to consider the task of speech enhancement as a holistic endeavor, and present a universal speech enhancement system that tackles 55 different distortions at the same time. Our approach consists of a generative model that employs score-based diffusion, together with a multi-resolution conditioning network that performs enhancement with mixture density networks. We show that this approach significantly outperforms the state of the art in a subjective test performed by expert listeners. We also show that it achieves competitive objective scores with just 4-8 diffusion steps, despite not considering any particular strategy for fast sampling. We hope that both our methodology and technical contributions encourage researchers and practitioners to adopt a universal approach to speech enhancement, possibly framing it as a generative task.

Speech anonymization has focused on the acoustic signal, leaving open how much speaker information the transcribed words themselves retain. In this work, we study pairwise speaker and demographic-attribute verification from transcribed speech. We introduce SPARC (Spoken Profile Authorship Recognition Corpus), a benchmark built from two podcast corpora comprising over 12,500 speakers, 1.5M quality-filtered utterances, and speaker-level inferred or source-provided labels for sex, age, and accent, and evaluate five attribution models and three LLMs on identity- and attribute-verification trials with episode- and metadata-based topic proxies. LLM adversaries distinguish same-speaker pairs from text alone. Attribute-label agreement is harder to verify, but LLMs perform sex- and accent-label verification above chance. Finally, releasing only short probe texts (50 words) degrades the strongest adversaries the most, offering a partial reduction in observed linkage.

Target speaker extraction (TSE) aims to isolate a specific speaker's speech from a mixture using speaker enrollment as a reference. While most existing approaches are discriminative, recent generative methods for TSE achieve strong results. However, generative methods for TSE remain underexplored, with most existing approaches relying on complex pipelines and pretrained components, leading to computational overhead. In this work, we present FlowTSE, a simple yet effective TSE approach based on conditional flow matching. Our model receives an enrollment audio sample and a mixed speech signal, both represented as mel-spectrograms, with the objective of extracting the target speaker's clean speech. Furthermore, for tasks where phase reconstruction is crucial, we propose a novel vocoder conditioned on the complex STFT of the mixed signal, enabling improved phase estimation. Experimental results on standard TSE benchmarks show that FlowTSE matches or outperforms strong baselines.

Conventional music structure analysis algorithms aim to divide a song into segments and to group them with abstract labels (e.g., 'A', 'B', and 'C'). However, explicitly identifying the function of each segment (e.g., 'verse' or 'chorus') is rarely attempted, but has many applications. We introduce a multi-task deep learning framework to model these structural semantic labels directly from audio by estimating "verseness," "chorusness," and so forth, as a function of time. We propose a 7-class taxonomy (i.e., intro, verse, chorus, bridge, outro, instrumental, and silence) and provide rules to consolidate annotations from four disparate datasets. We also propose to use a spectral-temporal Transformer-based model, called SpecTNT, which can be trained with an additional connectionist temporal localization (CTL) loss. In cross-dataset evaluations using four public datasets, we demonstrate the effectiveness of the SpecTNT model and CTL loss, and obtain strong results overall: the proposed system outperforms state-of-the-art chorus-detection and boundary-detection methods at detecting choruses and boundaries, respectively.

Text-to-music systems produce increasingly convincing audio, yet evaluation reveals little about whether the result matches user intent. A global text-audio relevance score can overlook the implicit intent in underspecified prompts and mask failures in specific requirements, such as instrumentation, structure, rhythm, or mood progression. To bridge this gap, we formulate text-to-music intent alignment as satisfying a per-request rubric of independently verifiable items covering both a request's explicit requirements and its implied musical intent. Scoring items individually makes evaluation diagnostic by intent source and musical dimension, rather than a single opaque score. We instantiate this as MuRA-Bench, a benchmark of real-world platform requests curated by music experts. We further propose MIRA (Musical Intent Refinement Agent), a test-time agent that first grounds a request's intent into rubrics, then searches over prompt revisions for a black-box generator under a bounded budget, iteratively generating music, verifying it against the rubrics, and using this feedback to guide a trajectory-aware tree search. Experiments across open-source and commercial backends show that MIRA improves intent alignment, enabling an open-source generator to achieve performance comparable to representative commercial systems (e.g. Suno and Mureka). Project page: https://mirareview.github.io/.

We introduce an approach to identifying speaker names in dialogue transcripts, a crucial task for enhancing content accessibility and searchability in digital media archives. Despite the advancements in speech recognition, the task of text-based speaker identification (SpeakerID) has received limited attention, lacking large-scale, diverse datasets for effective model training. Addressing these gaps, we present a novel, large-scale dataset derived from the MediaSum corpus, encompassing transcripts from a wide range of media sources. We propose novel transformer-based models tailored for SpeakerID, leveraging contextual cues within dialogues to accurately attribute speaker names. Through extensive experiments, our best model achieves a great precision of 80.3\%, setting a new benchmark for SpeakerID. The data and code are publicly available here: https://github.com/adobe-research/speaker-identification

Sound events are entities with semantic identities, locations, and trajectories, but current audio-language models usually reason about clips as global event content. Conversely, sound event localization models track source directions over time but offer limited semantic coverage for language reasoning. To address this gap, we introduce ST-AudioQA, a spatio-temporal audio QA dataset and benchmark built from first-order ambisonic (FOA) renderings of static and moving sound sources. Each scene provides source identity, activity, direction, distance, and motion metadata, enabling dense trajectory supervision and questions about what is sounding, where it is, how it moves, and how sources relate. We further propose ST-Audio Encoder, a time-resolved FOA audio encoder that learns event semantics together with source trajectories, and ST-AudioLM, which connects the audio tokens from the encoder to an LLM for spatio-temporal audio QA. Experiments show that this representation improves the semantic-localization tradeoff and yields stronger reasoning performance than static spatial and localization-oriented baselines.

Modern Text-to-Speech (TTS) systems increasingly leverage Large Language Model (LLM) architectures to achieve scalable, high-fidelity, zero-shot generation. However, these systems typically rely on fixed-frame-rate acoustic tokenization, resulting in speech sequences that are significantly longer than, and asynchronous with their corresponding text. Beyond computational inefficiency, this sequence length disparity often triggers hallucinations in TTS and amplifies the modality gap in spoken language modeling (SLM). In this paper, we propose a novel tokenization scheme that establishes one-to-one synchronization between continuous acoustic features and text tokens, enabling unified, single-stream modeling within an LLM. We demonstrate that these synchronous tokens maintain high-fidelity audio reconstruction and can be effectively modeled in a latent space by a large language model with a flow matching head. Moreover, the ability to seamlessly toggle speech modality within the context enables text-only guidance--a technique that blends logits from text-only and text-speech modes to flexibly bridge the gap toward text-only LLM intelligence. Experimental results indicate that our approach achieves performance competitive with state-of-the-art TTS and SLM systems while virtually eliminating content hallucinations and preserving linguistic integrity, all at a significantly reduced inference cost.

Long-form audio analysis requires systems to localize and integrate evidence distributed across extended recordings. While existing work primarily retrieves semantic content through structured textual representations, many real-world queries depend on acoustic evidence that is better preserved in continuous representations or raw audio. We introduce HARP (Hybrid Audio Retrieval Pipeline), an agentic framework and benchmark for systematically studying retrieval and evidence representations in long-audio analysis. Hybrid retrieval combining keyword and vector search shows the most robust performance. When paired with both metadata and retrieved audio as evidence, average answer accuracy improves by around 10% and rationale accuracy by around 6% over single-modality retrieval and evidence. Fine-grained evaluation shows that answer accuracy alone overestimates system capability and that HARP mostly follows human performance trends across query types. These results highlight the importance of combining structured retrieval with flexible access to audio evidence and evaluating long-audio systems beyond answer accuracy.

While recent zero-shot text-to-speech (TTS) models have significantly improved speech quality and expressiveness, mainstream systems still suffer from issues related to speech-text alignment modeling: 1) models without explicit speech-text alignment modeling exhibit less robustness, especially for hard sentences in practical applications; 2) predefined alignment-based models suffer from naturalness constraints of forced alignments. This paper introduces MegaTTS 3, a TTS system featuring an innovative sparse alignment algorithm that guides the latent diffusion transformer (DiT). Specifically, we provide sparse alignment boundaries to MegaTTS 3 to reduce the difficulty of alignment without limiting the search space, thereby achieving high naturalness. Moreover, we employ a multi-condition classifier-free guidance strategy for accent intensity adjustment and adopt the piecewise rectified flow technique to accelerate the generation process. Experiments demonstrate that MegaTTS 3 achieves state-of-the-art zero-shot TTS speech quality and supports highly flexible control over accent intensity. Notably, our system can generate high-quality one-minute speech with only 8 sampling steps. Audio samples are available at https://sditdemo.github.io/sditdemo/.

We present ArtifactNet, a lightweight framework that detects AI-generated music by reframing the problem as forensic physics -- extracting and analyzing the physical artifacts that neural audio codecs inevitably imprint on generated audio. A bounded-mask UNet (ArtifactUNet, 3.6M parameters) extracts codec residuals from magnitude spectrograms, which are then decomposed via HPSS into 7-channel forensic features for classification by a compact CNN (0.4M parameters; 4.0M total). We introduce ArtifactBench, a multi-generator evaluation benchmark comprising 6,183 tracks (4,383 AI from 22 generators and 1,800 real from 6 diverse sources). Each track is tagged with bench_origin for fair zero-shot evaluation. On the unseen test partition (n=2,263), ArtifactNet achieves F1 = 0.9829 with FPR = 1.49%, compared to CLAM (F1 = 0.7576, FPR = 69.26%) and SpecTTTra (F1 = 0.7713, FPR = 19.43%) evaluated under identical conditions with published checkpoints. Codec-aware training (4-way WAV/MP3/AAC/Opus augmentation) further reduces cross-codec probability drift by 83% (Delta = 0.95 -> 0.16), resolving the primary codec-invariance failure mode. These results establish forensic physics -- direct extraction of codec-level artifacts -- as a more generalizable and parameter-efficient paradigm for AI music detection than representation learning, using 49x fewer parameters than CLAM and 4.8x fewer than SpecTTTra.

Recent Large Audio Language Models (LALMs) excel in understanding but often lack transparent reasoning. To address this "black-box" limitation, we organized the Audio Reasoning Challenge at Interspeech 2026, the first shared task dedicated to evaluating Chain-of-Thought (CoT) quality in the audio domain. The challenge introduced MMAR-Rubrics, a novel instance-level protocol assessing the factuality and logic of reasoning chains. Featured Single Model and Agent tracks, the competition attracting 156 teams from 18 countries and regions. Results show agent systems currently lead in reasoning quality, utilizing iterative tool orchestration and cross-modal analysis. Besides, single models are rapidly advancing via reinforcement learning and sophisticated data pipeline. We details the challenge design, methodology, and a comprehensive analysis of state-of-the-art systems, providing new insights for explainable audio intelligence.

Audio language models are designed to understand speech, yet it remains unclear whether they capture how something is said beyond what is said. We present a mechanistic analysis of paralinguistic information in four open source models, Whisper-large-v2, Qwen2-Audio-7B Instruct, Qwen2.5-Omni-7B, and Chroma-4B, using the Expresso dataset with controlled speaking styles. We combine centered kernel alignment, linear probing with leave one speaker out evaluation, open ended tone prediction, and a content prosody leakage metric to trace how style information moves from the audio encoder to the final output. All models strongly encode speaking style in the late encoder, that is, the top third of the audio encoder's layers, but this information is consistently degraded before reaching the output. The projector reshapes representation geometry without removing information, while decoders differ in how much style they preserve depending on architecture and training objective. At the output level, models fall into two behaviors. Some are content driven, where predictions depend mainly on text. Others are acoustic driven, where predictions vary with speaking style. The leakage metric quantifies this difference, and qualitative results confirm it. Overall, we identify a gap between what models encode and what they use, highlighting a key limitation in current audio language models.

Data-based and learning-based sound source localization (SSL) has shown promising results in challenging conditions, and is commonly set as a classification or a regression problem. Regression-based approaches have certain advantages over classification-based, such as continuous direction-of-arrival estimation of static and moving sources. However, multi-source scenarios require multiple regressors without a clear training strategy up-to-date, that does not rely on auxiliary information such as simultaneous sound classification. We investigate end-to-end training of such methods with a technique recently proposed for video object detectors, adapted to the SSL setting. A differentiable network is constructed that can be plugged to the output of the localizer to solve the optimal assignment between predictions and references, optimizing directly the popular CLEAR-MOT tracking metrics. Results indicate large improvements over directly optimizing mean squared errors, in terms of localization error, detection metrics, and tracking capabilities.

Universal Speech Enhancement (USE) aims to restore speech quality under diverse degradation conditions while preserving signal fidelity. Despite recent progress, key challenges in training target selection, the distortion--perception tradeoff, and data curation remain unresolved. In this work, we systematically address these three overlooked problems. First, we revisit the conventional practice of using early-reflected speech as the dereverberation target and show that it can degrade perceptual quality and downstream ASR performance. We instead demonstrate that time-shifted anechoic clean speech provides a superior learning target. Second, guided by the distortion--perception tradeoff theory, we propose a simple two-stage framework that achieves minimal distortion under a given level of perceptual quality. Third, we analyze the trade-off between training data scale and quality for USE, revealing that training on large uncurated corpora imposes a performance ceiling, as models struggle to remove subtle artifacts. Our method achieves state-of-the-art performance on the URGENT 2025 non-blind test set and exhibits strong language-agnostic generalization, making it effective for improving TTS training data. Code and models will be released upon acceptance.

Existing multi-speaker dialogue systems bind speakers to utterances through structured supervision: per-turn tags, multi-stream transcriptions, or learnable speaker embeddings. These systems operate within speech-only pipelines that produce clean vocal sequences without the ambient texture of real conversations. We take a different approach. Our method, ScenA, conditions a text-to-audio flow-matching foundation model, pretrained on large-scale in-the-wild data, directly on multiple reference voices and a free-form natural language prompt that describes an entire multi-speaker audio scene. Leveraging such a foundational model allows us to inherit its capacity for natural, non-studio audio: background noise, room acoustics, overlapping dialogue, and spontaneous paralinguistic events, while adding multi-speaker control without any per-turn structure. Concretely, reference latents are concatenated into the model's token sequence and distinguished by lightweight identity-aware positional encodings. However, we identify a critical obstacle to this approach: the Reference Shortcut. During training under standard noise schedules, the model can identify the matching reference by acoustic similarity to the noisy target, bypassing the text prompt entirely. We address this with a high-noise-biased timestep distribution that forces the model to rely on the text prompt for speaker assignment. We evaluate ScenA on the CoVoMix2-Dialogue benchmark, showing that it outperforms existing multi-speaker systems on speaker-binding metrics while generating rich conversational audio with overlapping speech, emotional vocalizations, and ambient sound. Our results demonstrate the advantage of using a general-purpose audio model conditioned on a free-form scene description, rather than passing structured dialog scripts through a speech-only pipeline.

Developing large audio language models (LMs) capable of understanding diverse spoken interactions is essential for accommodating the multimodal nature of human communication and can increase the accessibility of language technologies across different user populations. Recent work on audio LMs has primarily evaluated their performance on short audio segments, typically under 30 seconds, with limited exploration of long-form conversational speech segments that more closely reflect natural user interactions with these models. We introduce Brutally Long Audio Bench (BLAB), a challenging long-form audio benchmark that evaluates audio LMs on localization, duration estimation, emotion, and counting tasks using audio segments averaging 51 minutes in length. BLAB consists of 833+ hours of diverse, full-length audio clips, each paired with human-annotated, text-based natural language questions and answers. Our audio data were collected from permissively licensed sources and underwent a human-assisted filtering process to ensure task compliance. We evaluate six open-source and proprietary audio LMs on BLAB and find that all of them, including advanced models such as Gemini 2.0 Pro and GPT-4o, struggle with the tasks in BLAB. Our comprehensive analysis reveals key insights into the trade-offs between task difficulty and audio duration. In general, we find that audio LMs struggle with long-form speech, with performance declining as duration increases. They perform poorly on localization, temporal reasoning, counting, and struggle to understand non-phonemic information, relying more on prompts than audio content. BLAB serves as a challenging evaluation framework to develop audio LMs with robust long-form audio understanding capabilities.