每日 arXiv

音频 · 语音 · 音乐 · 声学

每日新爬取的 arXiv 论文，按评分排序。中文摘要帮你快速判断是否值得深读。

2026-07-21
日期2026-07-21
已评分—
均分—
最高—

Daily Papers — 2026-07-21

12 papers on audio, speech, music, and acoustics.

1. Transcription Policy as a Latent Variable: Activating Controllable Verbatim ASR with Word-Level Timing

Authors: Laurin Wagner, Mario Zusag, Bernhard Thallinger

Categories: cs.CL | Accepted at Interspeech 2026 long track Score: 7.47/10 (Obj:9 Id:8 Ind:6 Comp:6 Eff:8 Nov:8 Rep:6)

  • Strength: 仅训练 27 个 token embedding（冻结 764M 参数）即将 German disfluency F1 从 10% 提升至 79%，证明 verbatim 能力是 latent 的；full fine-tuning 在英德双语上全面超越 Whisper/Canary-1B/Reverb/CrisperWhisper，首次以 attention-based 方法在 disfluent speech 上超越 forced a。
  • Weakness: 训练时间戳标签来自 MFA（与被超越的 baseline 同类），intended transcripts 由 GPT-4o 生成（闭源依赖，无独立人类校验），跨语言迁移仅验证 German（与 English 类型学接近），German 评测集自建且仅 202 句。
  • Full review: Claude Code 全文七公理审稿

Modern ASR models trained on heterogeneously annotated data treat transcription style (verbatim vs. intended) as an uncontrolled latent variable, causing measurable decoding instability, evaluation confounding (up to 60% of reported WER attributable to style mismatch), and unreliable word-level timing. We show that models already encode both styles; the challenge is controlled activation. Using coverage-aware decoder task tokens trained on parallel verbatim/intended transcript pairs, we raise German disfluency F1 from 10% to 79% zero-shot, despite English-only training. Full English-only fine-tuning surpasses all baselines in verbatim accuracy, disfluency detection, and intended-mode quality across both languages. We further introduce supervised cross-attention fine-tuning that improves word-level timestamps on disfluent speech beyond forced-alignment baselines. Finally, we propose verbatimize, a new task enabling scalable creation and enrichment of speech corpora with high-quality canonical verbatim transcriptions.

2. Content is What Remains: Invariant Speech Tokenization from Parallel Utterances

Authors: Laurin Wagner, Bernhard Thallinger, Miroslav Stankovic, Mario Zusag

Categories: cs.CL | Accepted at Interspeech 2026 Score: 6.89/10 (Obj:8 Id:5 Ind:8 Comp:8 Eff:6 Nov:7 Rep:8)

  • Strength: PINT 通过 parallel utterance alignment 将 speaker probe 从 93.1% 降至 1.2%，LM perplexity 降低 27-30%，且 RLE 压缩达 152 bits/s（接近文本 BPE 48 bits/s），invariance→compressibility 链条验证清晰。
  • Weakness: 未与 ContentVec、DC-Spin、StableToken、NAST 等目标高度重叠的 SOTA 方法直接比较，且缺少”微调但不加 alignment loss”的关键消融，无法确认提升归因于 parallel alignment 而非单纯微调+更多数据。
  • Full review: Claude Code 全文七公理审稿

Discrete speech tokenizers aim to disentangle semantic from acoustic information, yet targets from self-supervised learning (SSL) models like HuBERT retain non-linguistic variation: speaker identity, prosody, and channel conditions leak into the tokens, inflating entropy. Our key insight is that when enough speakers utter the same words under varying conditions, linguistic content is the only shared factor. We propose PINT (Parallel INvariant Tokenization), which fine-tunes an SSL encoder with alignment losses across parallel utterances and augmentations to distill this shared residual. PINT collapses identical words onto consistent token sequences, drastically reducing conditional entropy. Unlike ASR text, PINT tokens preserve frame-level temporal grounding and serve as drop-in semantic targets for audio codecs. Experiments show a 98.7% relative reduction in speaker probe accuracy (93.1% to 1.2%), a 42% lower ABX error rate, and 27-30% lower LM perplexity versus baselines, confirming that the right invariance is key to efficient learning.

3. Fusion Embedding: A Unified Embedding Space for Text, Image, Video, and Audio

Authors: Abdul Basit Tonmoy, Kazi Fardinul Hoque, Md. Shahrier Islam Arham, Arman Luthra

Categories: cs.CL, cs.SD | 23 pages, 5 figures. Models: https://huggingface.co/EximiusLabs/fusion-embedding-1-2b-preview and https://huggingface.co/EximiusLabs/fusion-embedding-2-2b-preview. Code: https://github.com/Eximius-Labs/fusion-embedding Score: 6.84/10 (Obj:8 Id:8 Ind:8 Comp:6 Eff:6 Nov:5 Rep:9)

  • Strength: 在冻结已部署 VL 嵌入基座上以 16.4M-60.6M 参数和单 GPU 数小时训练获得音频能力，bitwise base invariance 经逐位验证，emergent audio→image retrieval 达 29× chance（R@10 0.418），三个受控负面结果系统性地将 headroom 定位到 frozen LM 的音频 token 处理能力。
  • Weakness: 遗漏 e5-omni（ACL 2026）和 Conan-embedding-v3 两篇同类 omni-modal embedding 工作，绝对音频-文本检索 R@10 落后全量训练 CLAP 16-19 点（0.741 vs 0.906-0.928），跨模态评测仅限 VGGSound-696 单一 benchmark。
  • Full review: Claude Code 全文七公理审稿

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.

4. End-to-End Markov State Sequence Learning for Auditory Attention Decoding

Authors: Yushan Yashengjiang, Jie Zhang, Miao Sun, Huadong Liang, Xin Li et al.

Categories: cs.SD, cs.HC Score: 6.42/10 (Obj:8 Id:8 Ind:8 Comp:6 Eff:6 Nov:5 Rep:8)

  • Strength: 精心设计的消融实验（表 III）通过固定转换率控制，将序列感知发射学习（+8.5% 因果准确率）与转换率适配（+0.5%）分离开来，证明 CRF 训练塑造了发射表示本身，而非仅仅平滑输出。
  • Weakness: 遗漏了密切相关的先前工作 arXiv 2602.13447（Yao 等人，2026 年 2 月），该工作已经提出了在样本级进行集成马尔可夫模型 AAD 解码——早 5 个月，超出同期窗口——并且在 4s 窗口下，CRF 训练在 4 种主干网络中的 2 种上表现不如 HMM 后处理，且 23.3s 的切换延迟对于实时助听器不可用。
  • Full review: Claude Code 全文七公理审稿

Auditory attention decoding (AAD) identifies the speaker a listener attends to from neural responses like electroencephalography (EEG), making it a key algorithm in neuro-steered hearing aids. However, most neural AAD models are trained as independent short-window classifiers, despite auditory attention being a temporally persistent cognitive state and short-window EEG–audio evidence often being noisy and ambiguous. We propose an end-to-end Markov AAD framework based on conditional random field (CRF) that trains window-level neural emissions under a two-state attention prior. The framework treats the logits of any AAD backbone as Markov emissions, learns the transition rate from a standard HMM initialization, and jointly optimizes cross-entropy and CRF objectives, allowing temporal continuity to guide representation learning rather than merely smoothing predictions after training. We also introduce ESCNet, an EEG–speech correlation backbone that preserves time-aligned features and converts the difference between two mean Pearson correlations into state logits. We evaluate the framework with four emission backbones spanning correlation-based, convolutional, recurrent, and attention-based designs. On the dynamic AVGC dataset, CRF training generally outperforms post-hoc HMM smoothing; with ESCNet, it achieves $86.5\%$ causal and $92.4\%$ non-causal accuracy using $1$s windows. On the static KUL and USTC datasets, it improves causal decoding over fixed-rate post-hoc HMM baselines by $5.6\%$ and $2.0\%$, respectively, showing the superiority of learning AAD as attention state sequence over isolated-window classification.

5. Cyclic-Prefix-Free OFDM With Tail-Reuse Reconstruction for Distributed Acoustic Sensing

Authors: Ziang Chen, Zhiyang Xue, Zhongxing Tian, Zeyu Feng, Dongdong Zou et al.

Categories: eess.SP, physics.optics Score: 6.42/10 (Obj:8 Id:6 Ind:8 Comp:7 Eff:6 Nov:6 Rep:5)

  • Strength: 通过 tail-reuse 重构将 OFDM-DAS 的慢时间 Nyquist 极限从 4.69 kHz 提升到 9.16 kHz（约 2×），理论与仿真一致，盲定位实验在 5.2 km 光纤上验证了 500 Hz 和 3 kHz 振动恢复。
  • Weakness: 实验仅测试 2 个 PZT 频率且均在 CP-OFDM Nyquist 以下，未展示高频优势；未与任何非 OFDM DAS 带宽扩展技术比较；核心概念在雷达/通信领域已有先例但未引用讨论。
  • Full review: Claude Code 全文七公理审稿

Orthogonal frequency-division multiplexing (OFDM) enables frequency-domain reconstruction of the distributed Rayleigh backscatter channel in coherent distributed acoustic sensing (DAS), but an explicitly transmitted cyclic prefix (CP) lengthens the probing period and reduces the slow-time Nyquist limit of each range cell. We investigate a repeated cyclic-prefix-free OFDM waveform for DAS, in which the tail of the preceding useful block serves as a virtual cyclic extension. A finite-memory range condition for tail-reuse reconstruction is derived, and circular folding is identified when the useful period is shorter than the channel memory. For fixed useful-block length and fiber memory, removing the explicit CP increases the period-limited highest unaliased vibration frequency without changing the occupied-bandwidth-limited spatial resolution. In a 5.2-km numerical configuration, a 75-MSa/s processing rate and a 4096-sample useful block give a 54.61-us probing period and a 9.16-kHz slow-time Nyquist limit. Simulations verify tail reuse, the predicted folding boundary, and recovery of 100 vibration events from 800 Hz to 8800 Hz. Bandwidth-scaling simulations further show that joint processing of fine spatial observations improves differential-phase reliability and reconstruction SNR at a fixed reporting interval. Experiments on a 5.2-km coherent DAS link with 111.984-MHz occupied OFDM bandwidth blindly localize 500-Hz and 3-kHz PZT-induced vibrations at 5063.7 m and 5070.1 m, respectively, and recover their waveforms and spectra. The results demonstrate feasible tail-reuse channel reconstruction and quantify the extension of the unaliased vibration bandwidth.

6. Summary of DCASE 2026 Task 5: Audio-Dependent Question Answering

Authors: Haolin He, Renhe Sun, Zheqi Dai, Xingjian Du, Chunyat Wu et al.

Categories: eess.AS Score: 6.21/10 (Obj:8 Id:6 Ind:6 Comp:6 Eff:7 Nov:5 Rep:6)

  • Strength: 组织了 14 支队伍/36 份提交的首届音频依赖问答挑战赛，发现 233 个全失败题目和 96B vs 8B 仅 1.03pp 差距，提供清晰的社区改进目标。
  • Weakness: 核心贡献（ADQA/ADF/ADQA-Bench）来自 [1] 而非本文，且组织者 baseline 仅在 dev set 评分、eval set 缺失 baseline 对比，ADF 用 LALM 筛选又被 LALM 评测存在循环依赖。
  • Full review: Claude Code 全文七公理审稿

DCASE~2026 Task~5 introduces Audio-Dependent Question Answering (ADQA), which tests whether large audio-language models answer from the audio rather than from textual priors. An Audio-Dependency Filtering (ADF) pipeline combines silent-audio probing, per-option perplexity, a large language model (LLM) commonsense check, and human review to remove items solvable from text alone. The 3000 items that pass form the ADQA-Bench evaluation set, spanning music, speech, and environmental audio. The inaugural edition draws 14 teams and 36 submissions across two tracks defined by total parameter count (up to 100B and under 10B). A Chung-Ang University ensemble of MOSS-Audio-8B-Thinking and Qwen3-Omni-30B reaches the top overall accuracy at \pct{58.33}, and a MOSS-only configuration from the same team leads the sub-10B track at \pct{57.30}. Across the 30 submissions with a comparable development score, evaluation accuracy falls by 11.91 percentage points (pp) on average (median 10.91\,pp) on the hidden evaluation split, which is designed to be harder than the development split. The most common building blocks are: the MOSS-Audio-8B-Thinking backbone (13 of 36 submissions), Low-Rank Adaptation (LoRA) fine-tuning on AudioMCQ-StrongAC, and preference or reinforcement-learning objectives – Group Relative Policy Optimization (GRPO) in five teams, Group reward-Decoupled Normalization Policy Optimization (GDPO) in two. At test time, prompt engineering is near-universal, and majority or choice-permutation voting is common. Every system misses the same set of 233 evaluation items.

7. Towards Array-Invariant Speech Enhancement via Geometry-Aware Dynamic Convolution

Authors: Zhenglong Liu, Wangyou Zhang, Chenda Li, Yanmin Qian

Categories: eess.AS, cs.SD | Comments: 5 pages, 1 figure, 2 tables. Accepted at Interspeech 2026 Score: 5.60/10 (Obj:8 Id:5 Ind:8 Comp:5 Eff:6 Nov:5 Rep:5)

  • Strength: Geo-DConv 以极小参数开销 (+1.2-8.3%) 将固定阵列 SE 模型转化为阵列不变模型，在 RealMAN 上 SDR 达 9.72 dB (SpatialNet)，显著超越 FaSNet-TAC (+3.96 dB) 和 USES2-comp (+1.10 dB)，并在跨数据集 CHiME-4 (6-mic) 上泛化良好 (OVRL 2.64/2.73 vs 未处理 1.42)。
  • Weakness: 遗漏直接竞争对手 DeFTAN-AA (Interspeech 2024, 同任务) 作为 baseline，缺少组件级 ablation (TACT vs 简单 MLP, 动态卷积 vs 仿射变换) 和训练超参数/阵列采样分布的完整报告，Random 4-mic 设置下 SI-SDR/PESQ 退化未讨论。
  • Full review: Claude Code 全文七公理审稿

Multi-channel speech enhancement (SE) systems exhibit superior performance over single-channel methods but are constrained to fixed microphone array configurations. This restricts their real-world deployment across devices with diverse array geometries. While recent array-agnostic SE methods address variable microphone numbers and permutations, they largely fail to exploit explicit array geometry priors when available, missing a crucial cue for optimal spatial filtering. A Geometry-Aware Dynamic Convolution (Geo-DConv) framework is proposed, which explicitly leverages microphone coordinates to transform standard fixed-array SE models into robust array-invariant systems. Experiments are conducted on the recent real-recorded RealMAN multi-channel speech dataset. Results demonstrate that the proposed architecture enables two widely used fixed-array models to adapt to array-invariant settings, with consistent performance improvements across diverse array topologies.

8. CS-ETS: Chaos-Inspired Samba-Based EMG-To-Speech Synthesis with Nonlinear Chaotic Losses

Authors: Sajid Fardin Dipto, Tarikul Islam Tamiti, David Vergano, Luke Baja-Ricketts, Anomadarshi Barua

Categories: cs.SD, eess.SP Score: 5.05/10 (Obj:6 Id:5 Ind:8 Comp:5 Eff:5 Nov:4 Rep:6)

  • Strength: 在 ETS 任务上首次引入 chaos-inspired loss (LER+MSDFA)，配合 modified Samba encoder 实现 40.79% 参数压缩 (32M vs 54.1M)，并引入 PVA 方法使 ETS 领域首次能计算 LSD/STOI/PESQ 等 frame-level 指标。
  • Weakness: 核心混沌损失方法 (Lyapunov + DFA) 已在同组 CIS-BWE (ACL 2026) 中用于 speech BWE，方法新颖性实为任务迁移；缺失 Diff-ETS 和 emg2speech 等关键 2024-2025 baseline，且 encoder 架构变化与 loss 贡献未正交消融。
  • Full review: Claude Code 全文七公理审稿

We propose a chaos-inspired new architecture for EMG-to-Speech (ETS) synthesis called CS-ETS, which combines a Samba-based encoder with two novel chaos-inspired loss functions – Lyapunov Exponent Regularization (LER) and Multi-Scale Detrended Fluctuation Analysis (MSDFA). LER is designed based on Lyapunov exponents to capture nonlinear fluctuations and sensitivity to initial conditions. MSDFA exploits detrended fluctuation analysis to quantify fractal-like, long-range temporal chaotic correlation. CS-ETS surpasses prior work with a 40.79\% lower parameter count (32M vs 54.1M) and introduces a new Post-Vocoder Alignment approach that improves LSD by 2.1x, STOI by 4.7x, and SI-SDR by 1.25x. CS-ETS reduces computation by 13.33\% while maintaining improved performance. To the best of our knowledge, for the first time, we show how ETS can be supervised by the subtle non-linear chaotic physics with Samba attention to achieve a significantly smaller model with superior performance.

9. OmniReasoner: Thinking with Long Audio-Video via Native Tool Use

Authors: Yu Chen, Caorui Li, Ziyu Xiong, Yidong Wang, Mingqi Gao et al.

Categories: cs.CV Score: 4.70/10 (Obj:8 Id:5 Ind:5 Comp:4 Eff:4 Nov:2 Rep:8)

  • Strength: 将缩放工具范式从纯视频扩展到音视频全模态，配套 TimeAnchor 解决跨采样粒度时间一致性，并提供合成数据引擎避免人工区间标注，代码和数据开源。
  • Weakness: 核心贡献（zoom-in tool + SFT/RL 训练 + 粗到细推理）与 LongVT (CVPR 2026, 2025-11) 高度重叠，新颖性严重不足；缺失 LongVT、Zoom-Zero、OmniRAG-Agent 等直接竞品基线对比，效用无法独立验证。
  • Full review: Claude Code 全文七公理审稿

Long audio-video reasoning is difficult for omnimodal LLMs because the decisive evidence is often sparse, cross-modal, and too expensive to preserve with uniformly high-fidelity inputs. We introduce OmniReasoner, a tool-use post-training framework for Thinking with Long Audio-Video: omni-modal LLMs learn, via supervised fine-tuning and reinforcement learning, to decide whether and where to call a zoom-in tool before answering. OmniReasoner first builds a low-cost global preview of the full stream and then, when needed, calls the zoom-in tool with a requested temporal interval for higher-fidelity visual and audio inspection before answering. Because the model observes different sampling granularities before and after this call – a sparse global preview and a denser local clip – we introduce TimeAnchor, which keeps the tool’s temporal argument valid and round-trip-consistent across these granularities, rather than tied to frame indices from a particular sampling rate. To make this tool-use behavior trainable without expensive manual interval annotation, we build a Temporal Augmented Data Engine that synthesizes tool-use post-training trajectories by video editing and composition. Experiments across omnimodal and video benchmarks show that OmniReasoner improves both answer accuracy and temporal grounding while concentrating high-fidelity computation on informative regions. Code is available at https://github.com/RockyChen0205/OmniReasoner.

10. Teleportation Game: Quantum Teleportation in Multi-Agent Systems for Interactive Music

Authors: Eduardo Reck Miranda, Scott Yeiichi Oshiro

Categories: quant-ph, cs.SD Score: 3.68/10 (Obj:5 Id:2 Ind:4 Comp:4 Eff:3 Nov:5 Rep:4)

  • Strength: 论文将 SQPAM+PKBSE 量子音乐编码与标准量子隐形传态协议组合，实现 2-3 agent 级联交互，并提出用参数化 RX/RZ 替代 Pauli correction 的”interpreter module”以连续调节 state fidelity（Section 5 的 100 trials 显示 θ→π 时 mean fidelity 上升、variance 下降）。
  • Weakness: 缺乏与直接前作 MIq (Kirke 2020) 的比较和引用，无最简经典 baseline 对比，仅 1 个 flute 样本且 simulator/hardware 结论矛盾（agent 2→3 mcorr 从 0.717 降至 0.290），核心指标 melodic correlation 在多对 agent 上低于 0.15 表明 teleportation 几乎未传递有效音乐信息。
  • Full review: Claude Code 全文七公理审稿

This paper introduces an interactive music system with quantum musical agents that communicate by teleporting quantum states to one another. Human performers interact in real time with agents whose melodic and rhythmic behaviours are encoded as quantum states using Single Qubit Probability Amplitude Modulation (SQPAM) and structured through Quantum Phase Estimation (QPE). Up to three agents are combined within a single quantum circuit, with directed communication via quantum teleportation. We are interested in supporting ambiguous, transformative interactions reminiscent of free Jazz improvisation. Therefore, rather than treating noise and decoherence as limitations, the system embraces NISQ-era constraints as creative affordances, framing agent communication as quantum whispers, that is, deliberate, musically expressive imperfections in state transfer. We provide demonstrations and analyses based on melodic correlation, pitch-set distance, and state fidelity, where a continuum between imitation and divergence can be observed. We developed a tunable interpretation method to assess how agents reinterpret teleported states. This work positions teleportation as a promising interaction mechanism for agent-based quantum computer music and outlines future directions toward distributed ensembles connected via the Quantum Internet.

11. Towards a reproducible cross-venue method for quantifying crowd noise in stadiums

Authors: Alejandro Osses, Bente Ackermans, Helmer Nuijens, Rick Scholte

Categories: eess.AS | 14 pages Score: 2.95/10 (Obj:6 Id:2 Ind:5 Comp:5 Eff:2 Nov:2 Rep:2)

  • Strength: 论文清晰揭示了 GWR peak dB(A) 测量方法的标准化缺陷，并合理地将 ISO 1996-1 的 LAF,max / far-field criterion 迁移到场馆场景，142.2 dB(A) 与 Larson Davis 831 的 143 dB(Z) 过载点可疑接近这一发现具有警示价值。
  • Weakness: 论文零实验验证（作者自述 “no test cases are shared”），且 Barnard 2015 Inter-Noise 已提出几乎相同的 “proposed technique for measuring and comparing crowd noise” 先前工作，核心新颖性不成立。
  • Full review: Claude Code 全文七公理审稿

Public claims about the loudest stadium have been based on an instantaneous peak dB(A) reading measured at a single point, as popularised by the Guinness World Records “loudest crowd roar” category. The current record dates from 2014, where a maximum level of 142.2 dB(A) was registered. While compelling, those measurements lack standardisation, omitting relevant information such as the specific instrument that was used, the usage of time weighting, and the number of measurement positions that were tested. This lack of information does not allow a well-founded scientific comparison across sport venues. This study proposes a measurement framework in which spatially distributed acoustic measurement is the recommended route for a representative cross-venue comparison, while a single-anchor measurement can only serve as a minimum reporting baseline when distributed measurements are not feasible.

12. What the Waveform Knows: Transparent-first Speech and Audio Intelligence with Caption Studio

Authors: Cheng Siong Chin, Jianhua Zhang, Mohan Venkateshkumar

Categories: cs.SD, cs.AI Score: 2.21/10 (Obj:4 Id:2 Ind:5 Comp:2 Eff:1 Nov:1 Rep:4)

  • Strength: 论文对 speech-only analysis 的局限性有诚实讨论（Section V-E, V-F），明确拒绝 deception scoring，并提出了 measured/derived/unavailable 的 provenance 标签设计思路，这对企业部署场景有文档规范参考价值。
  • Weakness: 论文未执行任何正式评测（WER/DER 实验明确承认 “still require that run”），无 baseline 对比，无 ablation，无代码开源，系统本质是 Whisper+pyannote+librosa+VADER 的标准工具链组装，transparency-first framework 仅为 metadata tagging 实践，且 free-search 发现多个同类开源项目已实现类似整合。
  • Full review: Claude Code 全文七公理审稿

Caption Studio is a transparency-first speech and audio intelligence platform that transforms spoken audio and video into structured, searchable content through automated transcription, speaker diarization, speech analytics, signal-level audio analysis, and subtitle generation. The system is built on a FastAPI backend with a real-time dashboard and adopts a three-layer architecture comprising (i) a transcription and diarization core based on Whisper-class automatic speech recognition and pyannote speaker diarization, (ii) an audio intelligence layer that extracts acoustic and linguistic features, including waveforms, spectrograms, pitch, speaking rate, silence, filler-word frequency, and sentiment, directly from the audio signal, and (iii) an integration layer that supports data export and downstream workflow integration. A principal contribution of this work is the transparency-first framework, in which every reported metric is explicitly identified as measured, derived, or unavailable, thereby improving the traceability, interpretability, and reliability of speech analytics. The paper presents the system architecture, benchmarking methodology, explainability and uncertainty framework, and key considerations for enterprise-scale deployment.