Oriel — AI & technology worth noticing, every day.

VibeVoice Technical Report

Paper · https://arxiv.org/abs/2508.19205

This report presents VibeVoice, a novel model designed to synthesize long-form speech with multiple speakers by employing next-token diffusion, which is a unified method for modeling continuous data by autoregressively generating latent vectors via diffusion. To enable this, we introduce a novel continuous speech tokenizer that, when compared to the popular Encodec model, improves data compression by 80 times while maintaining comparable performance. The tokenizer effectively preserves audio fidelity while significantly boosting computational efficiency for processing long sequences. Thus, 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.

First seen by Oriel on .

AI summary

VibeVoice is a next-token diffusion model for long-form, multi-speaker speech synthesis that supports up to 90 minutes and reportedly surpasses open-source and proprietary dialogue models.

  • The model autoregressively generates continuous latent vectors through diffusion, providing a unified method for modeling continuous data.
  • Its continuous speech tokenizer achieves 80 times greater compression than Encodec while maintaining comparable performance and preserving audio fidelity.
  • Within a 64K context window, VibeVoice can synthesize conversations lasting up to 90 minutes with a maximum of four speakers.

#speech synthesis #diffusion models #audio tokenization

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hugging Face Papers: 187, latest observation ; +6 over observed period since

Comments

Comment count reported by the source.

  • Hugging Face Papers: 10, latest observation ; 0 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Papers2508.19205https://arxiv.org/abs/2508.19205 Read discussion