GrainSpeech is a compact speech synthesis model that uses a fixed‑receptive‑field convolutional encoder to reduce pitch, energy, and duration prediction errors by 36.0%, 17.3%, and 3.4% respectively. It introduces a Mel‑specific gradient‑variance supervision that improves fine‑scale variation while avoiding quality degradation. With only 264.8K parameters, GrainSpeech achieves 17.9× real‑time Mel generation on a microcontroller and attains UTMOS scores comparable to much larger models, using less than 1.5% of their parameters.
arXiv:2606. 30700v1 Announce Type: cross Abstract: Self-supervised learning enables audio representations that transfer across domains and tasks.
arXiv:2605.29859v2 Announce Type: replace-cross Abstract: Recent speech language models rely on encoders that are optimized separately from autoregressive models. Since these encoders are unaware of...
TUTTI is a new pre‑training framework for audio‑to‑score transcription that uses a large, fully synthetic multi‑instrument dataset generated by a symbolic music model. The approach trains a standard Transformer encoder‑decoder on these synthetic audio‑score pairs, producing a stronger foundational representation than single‑instrument training. When fine‑tuned on real datasets, TUTTI surpasses prior methods, achieving state‑of‑the‑art results and demonstrating strong cross‑instrument transferability.
Generalizable Audio-to-Score (A2S) transcription is fundamentally constrained by the severe scarcity of high-quality, real-world paired data. Relying solely on existing human-annotated datasets often...
AudioTETRIS is a training‑free, two‑stage token elimination framework designed for large audio language models. It leverages the geometry of audio encoder representations as an acoustic prior to compute per‑token importance scores, which guide progressive top‑k token selection across decoder layers. Experiments on Qwen2‑Audio‑7B, Phi‑4‑Multimodal, and Audio Flamingo 3 show that AudioTETRIS retains near‑full performance while using only 25% of audio tokens and outperforms existing training‑free baselines.
arXiv:2606. 06907v1 Announce Type: cross Abstract: Large audio language models (LALMs) extend large language models with an audio encoder and large-scale audio data.
The paper introduces a compression framework for the Whisper automatic speech recognition model that jointly optimizes six deployment dimensions—model size, temporal resolution, encoder token stride, low‑rank adaptation capacity, weight precision, and sparsity pattern—using NSGA‑III. The optimization targets three objectives: word error rate, inference FLOPs, and memory footprint. Evaluating 1,680 configurations, the study identifies compression combinations that outperform single‑axis scaling and notes that 1:4 structured sparsity cannot maintain acceptable accuracy within the tested budgets.
The paper investigates token sparsification for audio transformers, evaluating various pruning and merging methods across multiple datasets and model sizes. It introduces a new approach that blends attention scores with spectral cues from token embeddings, showing that frequency-aware techniques outperform attention-only methods. The study finds that spectrogram patch intensity-based sparsification is particularly effective for longer audio inputs, underscoring the benefit of incorporating frequency-domain information into sparsification strategies.
Mizar is a 159.3‑million‑parameter audio‑language model designed for devices with limited memory and computation. It couples a compact CED‑Small audio encoder with SmolLM2‑135M via a frequency‑merging mapper and is trained in three stages—audio‑language alignment, audio‑dependent fine‑tuning, and post‑training—to improve performance on audio‑question tasks. Across five random seeds, Mizar outperforms all other sub‑200M‑parameter ALMs on MMAU, MMAR, and ADQA‑clean, achieving mean accuracies of 52.92%, 42.42%, and 36.02% respectively, while enabling local inference on a single CPU with an average latency of 1.09 seconds for MMAU questions.
arXiv:2606. 14791v1 Announce Type: cross Abstract: Self-supervised learning advances audio representation for multimedia analysis.
arXiv:2609.14231v1 Announce Type: cross Abstract: Controllable synthesis of nonverbal vocalizations (NVVs) is es- sential for natural and expressive speech, but remains challeng- ing due to their aco...
The paper introduces a memory‑efficient audio synthesis architecture used in Siri Expressive Voices, converting semantic audio tokens into high‑fidelity audio via a residual vector quantization (RVQ) pipeline that decouples temporal and depth processing. A single Diffusion Transformer‑style depth decoder generates all RVQ levels autoregressively, while causal sliding‑window attention with fixed‑window key‑value caching keeps memory usage constant regardless of sequence length. Deployed on Apple’s AMX, the system achieves real‑time performance (≈10 ms per step) with only 21 MB peak runtime memory and 329 MB of on‑device assets, enabling continuous streaming of 20–320 s of audio and improving Mean Opinion Scores by +0.28 overall and +0.42 on conversational speech compared to the previous on‑device TTS system.