惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

爱范儿
爱范儿
博客园_首页
U
Unit 42
Apple Machine Learning Research
Apple Machine Learning Research
云风的 BLOG
云风的 BLOG
MongoDB | Blog
MongoDB | Blog
美团技术团队
H
Help Net Security
G
Google Developers Blog
B
Blog RSS Feed
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
aimingoo的专栏
aimingoo的专栏
Google DeepMind News
Google DeepMind News
J
Java Code Geeks
M
MIT News - Artificial intelligence
腾讯CDC
IT之家
IT之家
Vercel News
Vercel News
C
Check Point Blog
博客园 - 三生石上(FineUI控件)
Last Week in AI
Last Week in AI
I
InfoQ
博客园 - 司徒正美
A
About on SuperTechFans

cs.SD updates on arXiv.org

Mutual Forcing: Dual-Mode Self-Evolution for Fast Autoregressive Audio-Video Character Generation WhisperPipe: A Resource-Efficient Streaming Architecture for Real-Time Automatic Speech Recognition Walking Through Uncertainty: An Empirical Study of Uncertainty Estimation for Audio-Aware Large Language Models PSP: An Interpretable Per-Dimension Accent Benchmark for Indic Text-to-Speech Praxy Voice: Voice-Prompt Recovery + BUPS for Commercial-Class Indic TTS from a Frozen Non-Indic Base at Zero Commercial-Training-Data Cost Korean aegyo speech shows systematic F1 increase to signal childlike qualities All That Glitters Is Not Audio: Rethinking Text Priors and Audio Reliance in Audio-Language Evaluation RAS: a Reliability Oriented Metric for Automatic Speech Recognition Speech Enhancement Based on Drifting Models HeadRouter: Dynamic Head-Weight Routing for Task-Adaptive Audio Token Pruning in Large Audio Language Models Hallo-Live: Real-Time Streaming Joint Audio-Video Avatar Generation with Asynchronous Dual-Stream and Human-Centric Preference Distillation Talker-T2AV: Joint Talking Audio-Video Generation with Autoregressive Diffusion Modeling Robust Audio-Text Retrieval via Cross-Modal Attention and Hybrid Loss Spectro-Temporal Modulation Representation Framework for Human-Imitated Speech Detection UniSonate: A Unified Model for Speech, Music, and Sound Effect Generation with Text Instructions Do LLM Decoders Listen Fairly? Benchmarking How Language Model Priors Shape Bias in Speech Recognition Materialistic RIR: Material Conditioned Realistic RIR Generation SpeechParaling-Bench: A Comprehensive Benchmark for Paralinguistic-Aware Speech Generation ONOTE: Benchmarking Omnimodal Notation Processing for Expert-level Music Intelligence From Image to Music Language: A Two-Stage Structure Decoding Approach for Complex Polyphonic OMR ATIR: Towards Audio-Text Interleaved Contextual Retrieval Enhancing Speaker Verification with Whispered Speech via Post-Processing Environmental Sound Deepfake Detection Using Deep-Learning Framework Towards Streaming Target Speaker Extraction via Chunk-wise Interleaved Splicing of Autoregressive Language Model BEAT: Tokenizing and Generating Symbolic Music by Uniform Temporal Steps Deep Supervised Contrastive Learning of Pitch Contours for Robust Pitch Accent Classification in Seoul Korean HalluAudio: A Comprehensive Benchmark for Hallucination Detection in Large Audio-Language Models UAF: A Unified Audio Front-end LLM for Full-Duplex Speech Interaction Voice of India: A Large-Scale Benchmark for Real-World Speech Recognition in India Tadabur: A Large-Scale Quran Audio Dataset
RDSinger: Reference-based Diffusion Network for Singing V...
Kehan Sui, Jinxu Xiang, Fang Jin · 2024-10-29 · via cs.SD updates on arXiv.org

Singing voice synthesis (SVS) aims to produce high-fidelity singing audio from music scores, requiring a detailed understanding of notes, pitch, and duration, unlike text-to-speech tasks. Although diffusion models have shown exceptional performance in various generative tasks like image and video creation, their application in SVS is hindered by time complexity and the challenge of capturing acoustic features, particularly during pitch transitions. Some networks learn from the prior distribution and use the compressed latent state as a better start in the diffusion model, but the denoising step doesn't consistently improve quality over the entire duration. We introduce RDSinger, a reference-based denoising diffusion network that generates high-quality audio for SVS tasks. Our approach is inspired by Animate Anyone, a diffusion image network that maintains intricate appearance features from reference images. RDSinger utilizes FastSpeech2 mel-spectrogram as a reference to mitigate denoising step artifacts. Additionally, existing models could be influenced by misleading information on the compressed latent state during pitch transitions. We address this issue by applying Gaussian blur on partial reference mel-spectrogram and adjusting loss weights in these regions. Extensive ablation studies demonstrate the efficiency of our method. Evaluations on OpenCpop, a Chinese singing dataset, show that RDSinger outperforms current state-of-the-art SVS methods in performance.