Vocal Synthesizer Wiki
in: Work in Progress, Characters, Software,
and 3 more

DiffSinger

Sign in to edit
🛠 This subject is work in progress.
Please bear with us while improvements are being made and assume good faith until the edits are complete.
For information on how to help, see the guidelines. More subjects categorized here. 		🛠


DiffSinger is an open-source, AI-powered singing voice synthesis (SVS) software developed by a research team consisting of Jinglin Liu (刘静林), Chengxi Li (李成蹊), Yi Ren (任意), Feiyang Chen, and Zhou Zhao.

Concept[]

DiffSinger was proposed initially to the arXiv free distribution service as an an acoustic model for singing voice synthesis based on the diffusion probabilistic model. DiffSinger is a parameterized Markov chain that iteratively converts the noise into mel-spectrogram conditioned on the music score. By implicitly optimizing variational bound, DiffSinger can be stably trained and generate realistic outputs. To further improve the voice quality and speed up inference, the team introduced a shallow diffusion mechanism to make better use of the prior knowledge learned by the simple loss. Specifically, DiffSinger starts generation at a shallow step smaller than the total number of diffusion steps, according to the intersection of the diffusion trajectories of the ground-truth mel-spectrogram and the one predicted by a simple mel-spectrogram decoder. Besides, the team proposed boundary prediction methods to locate the intersection and determine the shallow step adaptively. The evaluations conducted on a Chinese singing dataset demonstrate that DiffSinger outperforms state-of-the-art SVS work. Extensional experiments also prove the generalization of our methods on text-to-speech task (DiffSpeech).[1]

History[]

WIP

Requirements[]

  • OS: Windows, macOS or Linux

External links[]

References[]

Navigation[]

v • e
DiffSinger Software
 DiffSinger

• Akiyama Daichi • Alekz Cyber • Amami Hiya • Ameko Kero • Aphelia • Ashera Lyre • Awata Weak • Dayton • Dongfang Zhizi • Emi • Félix • Gakuto Deshimaru • GAO • Gatito • Gil Qing • Halle Ruiz • Hanabata Haru • Hikuine Kiiro • Hitori no Shizuka • Hoshino Hanami • Huan Xing • Kanade Shia • Koe Kugo • Kohaku Merry • KoN • Kuroki Kasori • Lee Jieum • Lem • Lenz Blair • Lín Lí • Luther • Mimosa • Mine Laru • MURIN • Namine Criss • Namine Ritsu • Natsume Yuuri • Nishiren Gard • No.7 • Ogata Masato • Peiton • POTG • Printto Magicbeat • Qing Lin • Qixuan • Raine Reizo • Raine Rena • Roland Kanami • RUBY AI • Samuel Katsune • Sean • Shark • Shigashi • Shirane Kan • Shirohebi Airi • Shímián • Tsugine Saiko • Umidaji • VÂN • WALTT • Xia Yezi • Yamada Riku • Yamine Renri • Yokune Ruko • Yura Sashoune • Zhibin Shiyu

CODA: KITANE • Yeonu • So Yoon • POND8
Fraloids: KYANN • CORNELIA • HIBYA BÉRYL • VAKYN • LEVIS • Floralys • Fanny • Kimberley • Kristine • Kao • Cassandre
JAE VOCAL PROJECT: Canary • Xiao Meihua
Lunai Project: Llane Crow • KEIRO • Nero • Monone • Amaboshi Cipher • Liam • M1NT0 • Albus Korde • Denka Shaa • Karasu Yuutsukoe • Kuro Bousuku • Gahata Meiji
NeoDIVA: Aurelian Silva • Alex Floarea • Irone Aka • WURMLO1D • Andromeda
Tasteloid: Sweet • Bitter • Sour • Salty • Umami • Pepper • Yuzu
Team BRAPA: Umidaji • Reo
TGM: TIGER • Leif • Canary • TRITON • DANROU • Miyo • Adachi Rei
UFR: LYSE • ALYS • LEORA

  • Note: This is not a comprehensive list and only covers DiffSinger models featured on Vocal Synthesizer Wiki. Due to user creation capabilities, the total number of DiffSinger libraries is indefinite.
Community content is available under CC-BY-SA unless otherwise noted.