AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

NaturalSpeech : End-to-End Text-to-Speech Synthesis With Human-Level Quality

Xu Tan, Jiawei Chen, Haohe Liu, Jian Cong, Chen Zhang, Yanqing Liu, Xi Wang, Yichong Leng, Yuanhao Yi, Lei He, Sheng Zhao, Tao Qin and 2 more

TPAMI 2024 · Image generation › Autoregressive and latent generation · approach: Neural text-to-speech synthesis · 299 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add phoneme pre-training (ablation)NaturalSpeech · LJSpeechCMOS: -0.09 CMOS drop when removedImproved
add differentiable durator (ablation)NaturalSpeech · LJSpeechCMOS: -0.12 CMOS drop when removedImproved
add bidirectional prior/posterior module (ablation)NaturalSpeech · LJSpeechCMOS: -0.09 CMOS drop when removedImproved
add memory mechanism in VAE (ablation)NaturalSpeech · LJSpeechCMOS: -0.06 CMOS drop when removedImproved
end-to-end NaturalSpeech instead of VITSTTS system · LJSpeechCMOS: -0.20 -> 0Improved
end-to-end NaturalSpeech instead of FastSpeech 2 + HiFiGANTTS system · LJSpeechMOS: 4.32 -> 4.56Improved
use phoneme pre-training in FastSpeech 2 pipelineFastSpeech 2 + HiFiGAN · LJSpeechCMOS: 0.12 CMOS gainImproved
use predicted rather than ground-truth pitch/durationFastSpeech 2 + HiFiGAN · LJSpeechCMOS: -0.14Hurt
use mel decoder output instead of ground-truth melFastSpeech 2 + HiFiGAN · LJSpeechCMOS: -0.15Hurt
use vocoder on ground-truth mel instead of recordingsHiFiGAN · LJSpeechCMOS: -0.04Hurt
reduce Grad-TTS inference steps from 1000 to 10Grad-TTS + HiFiGAN · LJSpeechRTF: 4.120 -> 0.082Improved
keep strange-rhythm recordings in evaluationNaturalSpeech · LJSpeechCMOS: -0.01 -> +0.09Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy