@inproceedings{huang-etal-2023-fastdiff,
title = "{F}ast{D}iff 2: Revisiting and Incorporating {GAN}s and Diffusion Models in High-Fidelity Speech Synthesis",
author = "Huang, Rongjie and
Ren, Yi and
Jiang, Ziyue and
Cui, Chenye and
Liu, Jinglin and
Zhao, Zhou",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-acl.437",
doi = "10.18653/v1/2023.findings-acl.437",
pages = "6994--7009",
abstract = "Generative adversarial networks (GANs) and denoising diffusion probabilistic models (DDPMs) have recently achieved impressive performances in image and audio synthesis. After revisiting their success in conditional speech synthesis, we find that 1) GANs sacrifice sample diversity for quality and speed, 2) diffusion models exhibit outperformed sample quality and diversity at a high computational cost, where achieving high-quality, fast, and diverse speech synthesis challenges all neural synthesizers. In this work, we propose to converge advantages from GANs and diffusion models by incorporating both classes, introducing dual-empowered modeling perspectives: 1) FastDiff 2 (DiffGAN), a diffusion model whose denoising process is parametrized by conditional GANs, and the non-Gaussian denoising distribution makes it much more stable to implement the reverse process with large steps sizes; and 2) FastDiff 2 (GANDiff), a generative adversarial network whose forward process is constructed by multiple denoising diffusion iterations, which exhibits better sample diversity than traditional GANs. Experimental results show that both variants enjoy an efficient 4-step sampling process and demonstrate superior sample quality and diversity. Audio samples are available at \url{https://RevisitSpeech.github.io/}",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2023-fastdiff">
<titleInfo>
<title>FastDiff 2: Revisiting and Incorporating GANs and Diffusion Models in High-Fidelity Speech Synthesis</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rongjie</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="family">Ren</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ziyue</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chenye</namePart>
<namePart type="family">Cui</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinglin</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhou</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2023</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jordan</namePart>
<namePart type="family">Boyd-Graber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naoaki</namePart>
<namePart type="family">Okazaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Generative adversarial networks (GANs) and denoising diffusion probabilistic models (DDPMs) have recently achieved impressive performances in image and audio synthesis. After revisiting their success in conditional speech synthesis, we find that 1) GANs sacrifice sample diversity for quality and speed, 2) diffusion models exhibit outperformed sample quality and diversity at a high computational cost, where achieving high-quality, fast, and diverse speech synthesis challenges all neural synthesizers. In this work, we propose to converge advantages from GANs and diffusion models by incorporating both classes, introducing dual-empowered modeling perspectives: 1) FastDiff 2 (DiffGAN), a diffusion model whose denoising process is parametrized by conditional GANs, and the non-Gaussian denoising distribution makes it much more stable to implement the reverse process with large steps sizes; and 2) FastDiff 2 (GANDiff), a generative adversarial network whose forward process is constructed by multiple denoising diffusion iterations, which exhibits better sample diversity than traditional GANs. Experimental results show that both variants enjoy an efficient 4-step sampling process and demonstrate superior sample quality and diversity. Audio samples are available at https://RevisitSpeech.github.io/</abstract>
<identifier type="citekey">huang-etal-2023-fastdiff</identifier>
<identifier type="doi">10.18653/v1/2023.findings-acl.437</identifier>
<location>
<url>https://aclanthology.org/2023.findings-acl.437</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>6994</start>
<end>7009</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T FastDiff 2: Revisiting and Incorporating GANs and Diffusion Models in High-Fidelity Speech Synthesis
%A Huang, Rongjie
%A Ren, Yi
%A Jiang, Ziyue
%A Cui, Chenye
%A Liu, Jinglin
%A Zhao, Zhou
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Findings of the Association for Computational Linguistics: ACL 2023
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F huang-etal-2023-fastdiff
%X Generative adversarial networks (GANs) and denoising diffusion probabilistic models (DDPMs) have recently achieved impressive performances in image and audio synthesis. After revisiting their success in conditional speech synthesis, we find that 1) GANs sacrifice sample diversity for quality and speed, 2) diffusion models exhibit outperformed sample quality and diversity at a high computational cost, where achieving high-quality, fast, and diverse speech synthesis challenges all neural synthesizers. In this work, we propose to converge advantages from GANs and diffusion models by incorporating both classes, introducing dual-empowered modeling perspectives: 1) FastDiff 2 (DiffGAN), a diffusion model whose denoising process is parametrized by conditional GANs, and the non-Gaussian denoising distribution makes it much more stable to implement the reverse process with large steps sizes; and 2) FastDiff 2 (GANDiff), a generative adversarial network whose forward process is constructed by multiple denoising diffusion iterations, which exhibits better sample diversity than traditional GANs. Experimental results show that both variants enjoy an efficient 4-step sampling process and demonstrate superior sample quality and diversity. Audio samples are available at https://RevisitSpeech.github.io/
%R 10.18653/v1/2023.findings-acl.437
%U https://aclanthology.org/2023.findings-acl.437
%U https://doi.org/10.18653/v1/2023.findings-acl.437
%P 6994-7009
Markdown (Informal)
[FastDiff 2: Revisiting and Incorporating GANs and Diffusion Models in High-Fidelity Speech Synthesis](https://aclanthology.org/2023.findings-acl.437) (Huang et al., Findings 2023)
ACL