arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2408.13893cs.SDcs.CLeess.AS
收藏

SimpleSpeech 2:迈向基于流的标量潜在 Transformer 扩散模型的简单高效文本转语音

SimpleSpeech 2: Towards Simple and Efficient Text-to-Speech with Flow-based Scalar Latent Transformer Diffusion Models

Chinese University of Hong Kong(香港中文大学)

Dongchao Yang, Rongjie Huang, Yuanyuan Wang, Haohan Guo, Dading Chong, Songxiang Liu, Xixin Wu, Helen Meng

AI总结:

本文提出 SimpleSpeech 2,一个结合 AR 与 NAR 优势的简单高效非自回归 TTS 框架,通过引入流标量潜在 Transformer 扩散模型等改进,显著提升了语音生成的质量与速度,并支持多语言扩展。

摘要:

将文本转语音(TTS)扩展到大规模数据集已被证明是提升合成语音多样性与自然度的有效方法。宏观上,以往的大规模 TTS 模型可分为基于自回归(AR)的模型(如 VALL-E)和基于非自回归(NAR)的模型(如 NaturalSpeech 2/3)。尽管这些工作表现出良好的性能,但它们仍存在潜在弱点。例如,基于 AR 的模型受困于不稳定的生成质量和缓慢的生成速度;同时,一些基于 NAR 的模型需要音素级时长对齐信息,从而增加了数据预处理、模型设计和损失设计的复杂性。在本工作中,我们在先前发表的基础上,实现了一个简单高效的非自回归(NAR)TTS 框架,称为 SimpleSpeech 2。SimpleSpeech 2 有效结合了自回归(AR)和非自回归(NAR)方法的优势,提供以下关键优点:(1)简化的数据准备;(2)直观的模型与损失设计;(3)稳定、高质量的生成性能及快速的推理速度。与先前的发表相比,我们提出了(I)对语音分词器和噪声标签对 TTS 性能影响的详细分析;(II)四种不同类型的句子时长预测器;(III)一种新颖的基于流的标量潜在 Transformer 扩散模型。通过这些改进,我们展示了与先前工作及其他最先进(SOTA)的大规模 TTS 模型相比,在生成性能和生成速度上的显著提升。此外,我们展示了通过在多语言语音数据集上训练,SimpleSpeech 2 可以无缝扩展到多语言 TTS。演示可在:https://dongchaoyang.top/SimpleSpeech2_demo/ 获取。

英文摘要:

Scaling Text-to-speech (TTS) to large-scale datasets has been demonstrated as an effective method for improving the diversity and naturalness of synthesized speech. At the high level, previous large-scale TTS models can be categorized into either Auto-regressive (AR) based (\textit{e.g.}, VALL-E) or Non-auto-regressive (NAR) based models (\textit{e.g.}, NaturalSpeech 2/3). Although these works demonstrate good performance, they still have potential weaknesses. For instance, AR-based models are plagued by unstable generation quality and slow generation speed; meanwhile, some NAR-based models need phoneme-level duration alignment information, thereby increasing the complexity of data pre-processing, model design, and loss design. In this work, we build upon our previous publication by implementing a simple and efficient non-autoregressive (NAR) TTS framework, termed SimpleSpeech 2. SimpleSpeech 2 effectively combines the strengths of both autoregressive (AR) and non-autoregressive (NAR) methods, offering the following key advantages: (1) simplified data preparation; (2) straightforward model and loss design; and (3) stable, high-quality generation performance with fast inference speed. Compared to our previous publication, we present ({\romannumeral1}) a detailed analysis of the influence of speech tokenizer and noisy label for TTS performance; ({\romannumeral2}) four distinct types of sentence duration predictors; ({\romannumeral3}) a novel flow-based scalar latent transformer diffusion model. With these improvement, we show a significant improvement in generation performance and generation speed compared to our previous work and other state-of-the-art (SOTA) large-scale TTS models. Furthermore, we show that SimpleSpeech 2 can be seamlessly extended to multilingual TTS by training it on multilingual speech datasets. Demos are available on: {https://dongchaoyang.top/SimpleSpeech2\_demo/}.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑