Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.04500

VoiceWeaver: Staged Learning of Structured Controls for Expressive Speech and Sound-Event Generation

Xiaosu Su·Yun Cao·Yiping Ni·Xiaowei Yi
View original source ↗Read source document ↗Export metadata ↓

Abstract

Adding expressive and environmental controls to a speech generator requires learning heterogeneous attributes without losing earlier capabilities. VoiceWeaver addresses this problem with structured label prefixes and staged emotion-tone-event training in a shared text-audio model. Replay-based distillation retains earlier predictions, while embedding decorrelation and attribute dropout regularize conditioning. Evaluation separates single-attribute correctness, joint emotion-event generation, and three-attribute correctness. Rounded to whole percentage points, emotion accuracy is 86% in Chinese and 83% in English. Chinese emotion-event joint accuracy is about 70%, versus 56% for mixed-task training and 48% for Ming-omni-tts; English joint accuracy is 66%. Three-attribute accuracy is 59% in Chinese and 55% in English on 500 samples each (57% pooled). Text error rates exceed those of the external TTS baselines. Audio samples are available at https://anonymous.4open.science/api/repo/VoiceWeaver1-4D88/file/index.html.

Explore related subjects

cs.SD

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Xiaosu Su, Yun Cao, Yiping Ni, Xiaowei Yi. 2026-10-03. VoiceWeaver: Staged Learning of Structured Controls for Expressive Speech and Sound-Event Generation. https://arxiv.org/abs/2610.04500

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

A mathematical model of the vowel space

The articulatory-acoustic relationship is many-to-one and non linear and this is a great limitation for studying speech production. A simplification is proposed to set a bijection between the vowel space (f1, f2) and the parametric space of different vocal tract models. The generic area function model is based on mixtures of cosines allowing the generation of main vowels with two formulas. Then the mixture function is transformed into a coordination function able to deal with articulatory parameters. This is shown that the coordination function acts similarly with the Fant's model and with the 4-Tube DRM derived from the generic model.

cs.SD↗
arXiv2026

Re-purposing Multimodal Large Language Models for Audio-Text Retrieval

Audio-text retrieval is crucial for bridging acoustic signals and natural language. While contrastive dual-encoder architectures like CLAP have shown promise, they are fundamentally limited by the capacity of small-scale encoders. Specifically, the text encoders struggle to understand complex queries that require reasoning or world knowledge. In this paper, we propose AuroLA, a novel contrastive language-audio pre-trained model that re-purposes Multimodal Large Language Models (MLLMs) as a unified backbone for audio-text retrieval. Specifically, we make the following contributions: (i) we construct a scalable data pipeline that curates diverse audio from multiple sources and generates multi-granular captions, ranging from long descriptions to structured tags, via automated annotation; (ii) we adapt an MLLM for retrieval by prompting it to summarise the audio/text input and using the hidden state of a special token as audio/text embeddings. (iii) extensive experiments demonstrate that AuroLA consistently outperforms state-of-the-art dual-encoder models, including the recent PE-AV. This validates the effectiveness of MLLM as a unified backbone for audio-text retrieval.

cs.SD↗
arXiv2026

Rethinking Training Targets, Architectures and Data Quality for Universal Speech Enhancement

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. Model weights are available for download at: https://huggingface.co/nvidia/RE-USE.

cs.SD↗