Audio-Text Alignment

Forced aligners locate each word of a known transcript in the audio, and many speech pipelines depend on them. In our own data, a widely used CTC aligner inserts a pause at nearly every word boundary, and we found almost no aligner evaluated on telephone audio. We compare nine aligner configurations against human word boundaries on read, conversational and telephone English. The Montreal Forced Aligner (MFA) is the most precise, while CTC aligners consistently start words late, because most frames inside a word go to the blank token. We then fine-tune Whisper with LoRA to predict each word's start and end as timestamp tokens, training only on boundaries placed by people: 47 hours in 87 languages. The resulting aligner is more precise than MFA on every English test condition, and than MFA's own Korean and Dutch models. With FA-Bench's protocol and code on TIMIT, its word-boundary error is 11.6 ms, against 18.2 ms for the best published system. Two choices matter most on telephone conversations: placing the timestamp slots after the word they time, and training on noisy transcripts, which also makes the aligner nearly insensitive to errors in the transcript. The released model reaches 16.0 ms on real landline calls, against 36.5 ms for MFA. The models are available at https://huggingface.co/collections/Scicom-intl/force-alignment

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.

Audio language models pass speech and text through a single decoder. We ask whether that decoder represents a distinctive feature in the same direction when a phoneme is heard and when it is read. For minimal pairs of phonemes differing in one feature, we take the offset between the two members' mean representations. Averaging those offsets gives a direction for each stream, and we measure the cosine between the two. Because the two streams already agree about arbitrary phoneme pairs, we compare every measure against a reference built from random pairings rather than against zero. We apply this to 6 models, 7 features and 15 languages from 11 families. Only voicing in the two Qwen2.5-Omni models exceeds that reference after correction for multiple testing, and the reference varies by a factor of seven between models. In three of the six models, voicing has one direction in audio across the 14 languages with enough minimal pairs to measure it, and every language pair agrees in two of them. The model family, not the model size, predicts which stream represents a feature.

The Viterbi algorithm has been previously used to perform forced alignment of audio to text to mine training data from online resources. However, many existing implementations have quadratic time and space complexity, scaling poorly to long input sequences. We propose two optimizations to address this issue. First, we apply the Hirschberg algorithm to perform the alignment in place using linear memory. Second, we model the alignment between speech and text as a constrained random walk, allowing us to prune the search space with arbitrary confidence while accounting for transcription errors. The Hirschberg optimization reduces memory usage from 140 GB to 5 MB for three-hour inputs while producing identical alignments in one-third the time of torchaudio when both run on a CPU. We achieve an additional 2x speedup with pruning on inputs longer than 20 minutes while preserving alignment accuracy in more than 98% of tested cases.

Universal audio representations must preserve acoustic detail while making high-level concepts accessible across speech, music, environmental sound, and downstream models of different capacities. We study semantic refinement of an acoustically pretrained encoder by adding audio-description alignment to a foundation of BEST-RQ, reconstruction, and CTC. We compare matched control, shuffled-description, and correctly paired trajectories to distinguish correct correspondence from an extra contrastive objective. Each endpoint is frozen and evaluated with a temporal-mean linear probe and a sequence-aware LLM readout, testing whether the refined information is directly accessible and remains useful to a stronger model. Across three paired seeds, correct alignment improves domain-balanced classification by 4.66 points with the linear probe and 2.59 points with the sequence-aware LLM, with positive changes in every domain. Correct pairing accounts for 87% of the linear-probe gain, while the LLM shows its clearest correspondence-specific benefit in captioning. Dense acoustic objectives provide complementary gains under both readouts. A separate 24-layer continuation remains competitive with leading public encoders under the shared evaluator, supporting the recipe beyond the controlled study.

Speculative decoding speeds up generation by letting a cheap draft propose several tokens that a target model checks in one pass. In the single-model form, the draft is a lightweight module attached to the target rather than a separate model. Applying this design to Automatic Speech Recognition (ASR) introduces an extra problem. The draft can read the whole audio at every step, yet its proposals get worse as it runs on its own. Access is not localization. The accepted text keeps the transcript position explicit, but the draft must also track the changing audio position. In the primary matched comparison, per-step audio access changes the first proposal modestly but roughly doubles later-proposal acceptance. Fixed-width windows show that the audio position explains part of this gap. A correctly placed window recovers continuation, while an equally narrow window at the wrong position reduces it. Late-draft median error reaches 21 frames in the hardest reported condition, while target attention during verification stays within a 2-frame median. We test two ways to reduce this drift. The first reads the audio position from verification attention and uses it to guide the next draft round. It saves time only when the extra accepted tokens offset the readout cost. The second is AnchorDraft, which teaches the draft to track the audio position during training without changing the inference graph. The trained draft improves end-to-end speed at both tested target scales. These results show that ASR self-speculation depends on token prediction, audio-position tracking, and draft cost.

A single embedding space that covers text, images, video, and audio lets one index serve every query a user can pose. Embedding models built on vision-language backbones now lead text/image/video retrieval benchmarks but lack audio entirely, while audio-text retrieval is led by specialist systems that serve no other modality. We present the Fusion Embedding family, which adds audio to a frozen vision-language embedding base whose parameters are never updated: generation 1 (fusion-embedding-1) trains only a 16.4M-parameter connector between a frozen audio tower and the frozen base, and generation 2 (fusion-embedding-2) adds modality-gated deep adapters (44.2M parameters) whose branch never executes on text, image, or video inputs: their outputs are bit-for-bit those of the released base, verified after every training run. Because the base already binds text, images, and video, aligning audio to text alone makes audio-image retrieval emerge, with zero paired audio-visual training data. Alongside the recipe we map its design space with controlled negative results (rewriting training captions with an LLM, substituting a leaderboard-stronger audio tower, and widening the connector each reduce retrieval) and with training-protocol findings that we expect to transfer to any frozen decoder-LM embedding backbone. Both generations train in hours on a single GPU. Weights, code, and the evaluation harness are openly released.

Speech-to-text alignment means finding the temporal boundaries of each word in the audio. Some models provide such an alignment directly and others do not. Connectionist temporal classification (CTC) and transducer models have an alignment by construction, whereas attention-based encoder-decoders (AED) and speech large language models (LLMs) do not, and their word timings are usually read off the attention weights instead. All of these signals live on the encoder frame grid, which bounds their temporal precision. We study a generic gradient-based alignment that applies to any differentiable ASR model. We take the gradient of each teacher-forced token log probability with respect to the input, reduce it to a per-frame saliency, and decode the resulting matrix into word boundaries with a single dynamic-programming pass. The method needs no training, no model modification and no alignment heads, works across all model families including the speech LLMs, and aligns on the input grid rather than on the coarser encoder grid. We evaluate it on sixteen models from four families, on read (TIMIT) and spontaneous (Buckeye) speech, each against the model's own native or attention-based alignment. We find that the gradient yields a usable alignment for every model, that it is usually somewhat behind a strong native aligner but better where the native alignment is weak, as for the streaming models, and that its main disadvantage is the cost of one backward pass per token.

Aligner-Encoders are recently proposed seq2seq end-to-end ASR models that replace decoder attention by predicting the uth token directly from the u-th encoder position, so the encoder must learn the alignment internally without cross-attention or a transducer lattice. In practice, this alignment often forms abruptly in the upper layers, making training sensitive and brittle on long utterances. We propose InterAligner, which adds an intermediate Aligner objective so alignment can form progressively across depth, together with an intermediate CTC loss (InterCTC) to stabilize optimization. On LibriSpeech with a 17-layer Conformer, a final-only Aligner reaches 5.0/7.8 WER (test-clean/other). InterCTC improves to 3.4/6.0, and InterAligner further reduces WER to 3.1/5.6 with the largest gains on long utterances.

Acoustic environments often contain multiple overlapping sound events, and the same acoustic scene can be described using diverse textual expressions, making audio-text alignment inherently ambiguous. This paper proposes a probabilistic audio-language pretraining framework to model many-to-many correspondence ambiguity in audio-text alignment. Unlike conventional contrastive methods that learn deterministic point embeddings, our approach represents each modality as a distribution and learns uncertainty-aware cross-modal alignment. Rather than relying on masking-based uncertainty simulation, we mix audio-text pairs to create overlapping sounds that better reflect real acoustic mixtures and capture semantic inclusion relations among sound events. We further introduce a multi-level inclusion loss to enforce representations consistent with these relations. Experiments on audio-text retrieval benchmarks show that the proposed method outperforms deterministic baselines.

Text-to-audio (TTA) generation, synthesizing audio from natural language, has been widely studied for its ability to capture precise user intent. To effectively advance TTA models, it is essential to reliably evaluate generated audio without relying on costly human subjective ratings, motivating the development of automatic evaluation metrics that correlate well with human judgments. While recent CLAP-based metrics provide practical reference-free solutions, their coarse-grained text-audio similarity matching often correlates poorly with human ratings. To address this, we propose ELSA, a reference-free evaluation metric for fine-grained text-audio alignment. ELSA decomposes generated audio guided by distinct acoustic events derived from the text query and assesses event-level alignment. Experiments across four TTA benchmarks show that ELSA reveals a higher correlation with human subjective ratings than prior metrics, highlighting its effectiveness for reliable TTA evaluation.

State-of-the-art text-to-music generation systems rely on massive proprietary datasets and industrial-scale compute, making it impossible to disentangle architectural contributions from resource advantages. We propose score-aware training, which treats audio-caption alignment score as a direct supervision signal throughout the pipeline. Rather than discarding low-scoring segments, we repurpose them via a CLAP-conditioned Beta noise timestep schedule that routes them to high-noise training regimes, acting as an effective implicit regularizer. Complementarily, segment-level filtering removes the most misaligned examples, and a two-stage caption procedure bridges the distribution gap between verbose training captions and concise inference prompts. A REPA auxiliary loss further transfers structured semantic knowledge from pretrained CLAP and MuQ encoders without additional data. Our 450M-parameter FluxAudio-based system, submitted to the ICME 2026 ATTM Grand Challenge Efficiency Track, ranked 2nd across both tracks in the objective evaluation and 3rd in the Efficiency Track in the final MOS evaluation.

Multimodal Large Language Models (MLLMs) effectively integrate text and audio to interpret context in complex interactive dialogues. However, the internal mechanisms by which heterogeneous modalities influence model behavior remain opaque. While Shapley Values (SV) provide a robust, model-agnostic framework for local explainability in text-based NLP, their extension to multimodal data is hindered by cross-channel dependencies, intricate dialogue structures, and the prohibitive computational complexity of dense audio representations. In this work, we formalize a multimodal extension of the Shapley Value framework, treating discrete text tokens and aligned audio segments as cooperative features. To ensure computational feasibility, we deploy a suite of efficient estimation strategies: exact SV computation for low-dimensional inputs and sampling-based approximations - including Monte Carlo permutations and stratified sampling with Neyman-optimal allocation - to minimize variance under constrained computational budgets. To resolve the granularity mismatch between modalities, we propose Spectrogram-Guided Phonetic Alignment (SGPA), a novel preprocessing method that maps high-frequency audio streams to interpretable, word-aligned segments. Our contribution is twofold: first, we provide an open-source, model-agnostic Python package and a companion GUI for the computation and interactive visualization of multimodal attributions. Second, we evaluate our framework using curated subsets of the VoiceBench and Infinity Instruct datasets across diverse multilingual scenarios. Our experimental results reveal that input modality is a primary driver of attribution volatility and demonstrate that standard syntactic importance proxies often fail to predict model attention in multimodal, cross-lingual contexts.

Speech LLM post-training increasingly relies on efficient cross-modal alignment and robust low-resource adaptation, yet collecting large-scale audio-text pairs remains costly. Text-only alignment methods such as TASU reduce this burden by simulating CTC posteriors from transcripts, but they provide limited control over uncertainty and error rate, making curriculum design largely heuristic. We propose TASU2, a controllable CTC simulation framework that simulates CTC posterior distributions under a specified WER range, producing text-derived supervision that better matches the acoustic decoding interface. This enables principled post-training curricula that smoothly vary supervision difficulty without TTS. Across multiple source-to-target adaptation settings, TASU2 improves in-domain and out-of-domain recognition over TASU, and consistently outperforms strong baselines including text-only fine-tuning and TTS-based augmentation, while mitigating source-domain performance degradation.

The audio research community depends on open generative models as foundational tools for building novel approaches and establishing baselines. In this report, we present Woosh, Sony AI's publicly released sound effect foundation model, detailing its architecture, training process, and an evaluation against other popular open models. Being optimized for sound effects, we provide (1) a high-quality audio encoder/decoder model and (2) a text-audio alignment model for conditioning, together with (3) text-to-audio and (4) video-to-audio generative models. Distilled text-to-audio and video-to-audio models are also included in the release, allowing for low-resource operation and fast inference. Our evaluation on both public and private data shows competitive or better performance for each module when compared to existing open alternatives like StableAudio-Open and TangoFlux. Inference code and model weights are available at https://github.com/SonyResearch/Woosh. Demo samples can be found at https://sonyresearch.github.io/Woosh/.

Auscultation is a vital diagnostic tool, yet its utility is often limited by subjective interpretation. While general-purpose Audio-Language Models (ALMs) excel in general domains, they struggle with the nuances of physiological signals. We propose a framework that aligns multi-site auscultation recordings directly with a frozen Large Language Model (LLM) embedding space via gated cross-attention. By leveraging the LLM's latent world knowledge, our approach moves beyond isolated classification toward holistic, patient-level assessment. On the CaReSound benchmark, our model achieves a state-of-the-art 0.865 F1-macro and 0.952 BERTScore. We demonstrate that lightweight, domain-specific encoders rival large-scale ALMs and that multi-site aggregation provides spatial redundancy that mitigates temporal truncation. This alignment of medical acoustics with text foundations offers a scalable path for bridging signal processing and clinical assessment.

Speech-to-text alignment is a critical component of neural text to speech (TTS) models. Autoregressive TTS models typically use an attention mechanism to learn these alignments on-line, while non-autoregressive end to end TTS models rely on durations extracted from external sources. In this paper, we propose a novel duration prediction framework that can give promising phoneme-level duration distribution with given text. In our experiments, the proposed duration model has more precise prediction and adaptation ability to conditions, compared to previous baseline models. Specifically, it makes a considerable improvement on phoneme-level alignment accuracy and makes the performance of zero-shot TTS models more robust to the mismatch between prompt audio and input audio.