VibeVoice: Expressive Podcast Generation with Next-Token Diffusion

ICLROral2026

Authors: Zhiliang Peng, Jianwei Yu, Wenhui Wang, Yaoyao Chang, Yutao Sun, Li Dong, Yi Zhu, Weijiang Xu, Hangbo Bao, Zehua Wang, Shaohan Huang, Yan Xia, Furu Wei
Affiliation: Microsoft
Venue: ICLR 2026
Track: Oral

TL;DR

VibeVoice can synthesize long-form speech for up to 90 minutes (in a 64K context window length) with a maximum of 4 speakers, capturing the authentic conversational vibe and surpassing open-source and proprietary dialogue models.

Opening excerpt from the authors’ abstract. source

Read the paper

OpenReview

Topics

diffusion speech

← All ICLR 2026 Oral papers · Browse the whole archive