arXiv||1 min read
Error Rate is Not Enough: Costs of Augmentation in Low-Resource ASR
Focuses on Error Rate is Not Enough: Costs of Augmentation in Low-Resource ASR.
Topic archive
Audio AI news
Audio models, speech recognition, music, and sound generation. This page collects the latest briefings that match the topic so readers can follow one area without scanning the full feed.
Indexed briefings
533
Latest source-linked updates, ordered newest first.
Latest
Recent briefings
arXiv||1 min read
OmniCapBench: A Deep-Structured Evaluation Framework for Fine-Grained Audio-Visual Captioning
Focuses on OmniCapBench: A Deep-Structured Evaluation Framework for Fine-Grained Audio-Visual Captioning.
arXiv||1 min read
Steerspeech: Activation Steering For Emotion Control In Generated Speech
Focuses on Steerspeech: Activation Steering For Emotion Control In Generated Speech.
arXiv||1 min read
MIRA: A Musical Intent Refinement Agent for Aligning Text-to-Music Generation with User Intent
Focuses on MIRA: A Musical Intent Refinement Agent for Aligning Text-to-Music Generation with User Intent.
arXiv||1 min read
Real-Time Joint Audio-Video Generation by Parallel Adapter Composition
Focuses on Real-Time Joint Audio-Video Generation by Parallel Adapter Composition.
arXiv||1 min read
CrossEdit: Cross-Modal Training Enables Rich Audio-Visual Editing
Focuses on CrossEdit: Cross-Modal Training Enables Rich Audio-Visual Editing.
arXiv||1 min read
Benchmarking Adult Addressee Classification Across Child- and Adult-Directed Speech Datasets
Focuses on Benchmarking Adult Addressee Classification Across Child- and Adult-Directed Speech Datasets.
arXiv||1 min read
Steerspeech: Activation Steering For Emotion Control In Generated Speech
Focuses on Steerspeech: Activation Steering For Emotion Control In Generated Speech.
arXiv||1 min read
DuRe-ST: Dual-Relation Spectro-Temporal Modeling for Speech Deepfake Detection
Focuses on DuRe-ST: Dual-Relation Spectro-Temporal Modeling for Speech Deepfake Detection.
arXiv||1 min read
Source-Directed Trajectory Perturbation at First-Order Cost for Domain Generalization in Speech Deepfake Detection
Focuses on Source-Directed Trajectory Perturbation at First-Order Cost for Domain Generalization in Speech Deepfake Detection.
arXiv||1 min read
AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video Generation
Focuses on AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video Generation.
arXiv||1 min read
A Study on Improving Multi-class Audio Source Separation Via Decoupled CLAP Query Optimization and an Automated Data Engine
Focuses on A Study on Improving Multi-class Audio Source Separation Via Decoupled CLAP Query Optimization and an Automated Data Engine.
arXiv||1 min read
LIFT-SE: Linguistic Inference Followed by Flow Transformation for Generative Speech Enhancement
Focuses on LIFT-SE: Linguistic Inference Followed by Flow Transformation for Generative Speech Enhancement.
arXiv||1 min read
Itgan at NADI 2026 shared task: Parameter-Efficient Whisper Adaptation for Robust, Mixed-Dialect and Code-Switched Arabic ASR
Focuses on Itgan at NADI 2026 shared task: Parameter-Efficient Whisper Adaptation for Robust, Mixed-Dialect and Code-Switched Arabic ASR.
arXiv||1 min read
CARES: A Controlled Synthetic Benchmark of Speaker Reactions to Sound
Focuses on CARES: A Controlled Synthetic Benchmark of Speaker Reactions to Sound.
arXiv||1 min read
Training-Free Instruction TTS Gender Bias Calibration Using Model-Adaptive Steering
Focuses on Training-Free Instruction TTS Gender Bias Calibration Using Model-Adaptive Steering.
arXiv||1 min read
Direction-preserving Spatial Audio Signal Enhancement in The Spherical Harmonic Domain
Focuses on Direction-preserving Spatial Audio Signal Enhancement in The Spherical Harmonic Domain.
arXiv||1 min read
Tracing Inputs, Verifying Outputs: Validating Attribution in Music Generation
Focuses on Tracing Inputs, Verifying Outputs: Validating Attribution in Music Generation.
arXiv||1 min read
Mitigating Accent-Language Confusion in Self-Supervised Speech Representations for Language Identification
Focuses on Mitigating Accent-Language Confusion in Self-Supervised Speech Representations for Language Identification.
arXiv||1 min read
BanglaRhet: Benchmarking Classical and Transformer Models for Rhetorical and Persuasion Detection in Bangla Political Speech
Focuses on BanglaRhet: Benchmarking Classical and Transformer Models for Rhetorical and Persuasion Detection in Bangla Political Speech.
arXiv||1 min read
Beyond Token Revision: Investigating Mask-and-Replace Diffusion for Zero-Shot Text-to-Speech
Focuses on Beyond Token Revision: Investigating Mask-and-Replace Diffusion for Zero-Shot Text-to-Speech.
arXiv||1 min read
TiTok: Audio-Visual LLM for Multi-Segment Temporal Grounding
Focuses on TiTok: Audio-Visual LLM for Multi-Segment Temporal Grounding.
arXiv||1 min read
VM-ARRAYDPS: Virtual Microphone Augmented Diffusion Posterior Sampling for Unsupervised Blind Speech Separation
Focuses on VM-ARRAYDPS: Virtual Microphone Augmented Diffusion Posterior Sampling for Unsupervised Blind Speech Separation.
arXiv||1 min read
Dialect-Robust Speech Language Models with Synthetic Pseudo-Dialect Augmentation
Focuses on Dialect-Robust Speech Language Models with Synthetic Pseudo-Dialect Augmentation.
arXiv||1 min read
Lyapunov-Inspired LyRIC Activation and GLARE Attention in Chaos-Guided State Space Modeling for EMG-To-Speech (ETS) Synthesis
Focuses on Lyapunov-Inspired LyRIC Activation and GLARE Attention in Chaos-Guided State Space Modeling for EMG-To-Speech (ETS) Synthesis.
arXiv||1 min read
BanglaBox: A Phonetically-Balanced Corpus and Data-Efficient Foundation-Model Adaptation for Bangla Text-to-Speech with Zero-Shot Voice Cloning
Focuses on BanglaBox: A Phonetically-Balanced Corpus and Data-Efficient Foundation-Model Adaptation for Bangla Text-to-Speech with Zero-Shot Voice Cloning.
arXiv||1 min read
SwinDS-BWE: A Parameter-Efficient Swin-1D Lattice Generator with Decision-Science Discriminators for Speech Bandwidth Extension
Focuses on SwinDS-BWE: A Parameter-Efficient Swin-1D Lattice Generator with Decision-Science Discriminators for Speech Bandwidth Extension.
arXiv||1 min read
OmniSeek: Native Tool Integration for Multi-turn Audio-Visual Reasoning
Focuses on OmniSeek: Native Tool Integration for Multi-turn Audio-Visual Reasoning.
arXiv||1 min read
Shared-State Local Translations for Training-Free Voice Conversion
Focuses on Shared-State Local Translations for Training-Free Voice Conversion.
arXiv||1 min read
Removing Timing Shortcuts Improves Non-Invasive Brain-to-Text
Focuses on Removing Timing Shortcuts Improves Non-Invasive Brain-to-Text.
arXiv||1 min read
GLARE: Generating Listening Heads with Appropriate Reactions
Focuses on GLARE: Generating Listening Heads with Appropriate Reactions.
arXiv||1 min read
LAST: Looped Audio Spectrogram Transformer
Focuses on LAST: Looped Audio Spectrogram Transformer.
arXiv||1 min read
From Isolated Feature to Orbits: Discovering Music Concepts via Multi-SAE Alignment
Focuses on From Isolated Feature to Orbits: Discovering Music Concepts via Multi-SAE Alignment.
arXiv||1 min read
AVSD-Scenes: A Dataset for Audio-Visual Description of Urban Scenes
Focuses on AVSD-Scenes: A Dataset for Audio-Visual Description of Urban Scenes.
arXiv||1 min read
Beyond Decodability: Do Acoustic Factors Drive Predictions in Speech-Based Alzheimer's Assessment?
Focuses on Beyond Decodability: Do Acoustic Factors Drive Predictions in Speech-Based Alzheimer's Assessment?.
arXiv||1 min read
Teaching LLMs to Hear Who Spoke What: Metadata-Supervised Pretraining for Encoder-Free Speech-LLMs
Focuses on Teaching LLMs to Hear Who Spoke What: Metadata-Supervised Pretraining for Encoder-Free Speech-LLMs.