WaveGlow

E200567

autoregressive-free vocoder deep learning model flow-based generative model neural network model speech synthesis system text-to-speech model

WaveGlow is a flow-based generative neural network model for fast, high-quality text-to-speech audio synthesis.

Try in SPARQL Jump to: Surface forms Statements Referenced by

All labels observed (2)

Label	Occurrences
WaveGlow canonical	2
WaveGlow: A Flow-based Generative Network for Speech Synthesis	1

Statements (48)

Predicate	Object
instanceOf	autoregressive-free vocoder ⓘ deep learning model ⓘ flow-based generative model ⓘ neural network model ⓘ speech synthesis system ⓘ text-to-speech model ⓘ
advantageOverAutoregressiveModels	lower inference latency ⓘ parallel sampling ⓘ
architectureComponent	series of flow steps ⓘ upsampling network for conditioning ⓘ
audioQuality	near state-of-the-art at time of publication ⓘ
basedOn	Glow ⓘ WaveNet ⓘ
codeRepository	GitHub ⓘ
comparedWith	ClariNet ⓘ Parallel WaveNet ⓘ WaveNet ⓘ
designedTo	enable real-time TTS ⓘ replace autoregressive vocoders ⓘ
developedBy	NVIDIA Corporation ⓘ surface form: NVIDIA
distributionAssumption	simple prior distribution on latent space ⓘ
domain	audio generation ⓘ speech processing ⓘ
framework	PyTorch ⓘ
input	mel-spectrograms ⓘ
introducedAt	2018 ⓘ
language	Python ⓘ
output	time-domain audio waveform ⓘ
paperTitle	WaveGlow self-linksurface differs ⓘ surface form: WaveGlow: A Flow-based Generative Network for Speech Synthesis
probabilityModel	exact likelihood model ⓘ
property	fast parallel audio generation ⓘ fully convolutional architecture ⓘ high-quality speech synthesis ⓘ single-network architecture ⓘ
publisher	NVIDIA Corporation ⓘ surface form: NVIDIA
releasedAs	open source ⓘ
supports	GPU acceleration ⓘ
task	neural vocoding ⓘ text-to-speech synthesis ⓘ
trainingDataType	paired text and speech corpora ⓘ
trainingObjective	log-likelihood maximization ⓘ maximum likelihood ⓘ
usedFor	neural TTS systems ⓘ speech synthesis research ⓘ voice assistants ⓘ
uses	affine coupling layers ⓘ invertible 1x1 convolutions ⓘ normalizing flows ⓘ

Referenced by (3)

Full triples — surface form annotated when it differs from this entity's canonical label.

WaveNet → ledTo → WaveGlow ⓘ

WaveGlow → paperTitle → WaveGlow self-linksurface differs ⓘ

this entity surface form: WaveGlow: A Flow-based Generative Network for Speech Synthesis

Parallel WaveNet → relatedTo → WaveGlow ⓘ