Microsoft has built an open-source voice AI system, VibeVoice, that can turn text into long, realistic conversations lasting hours, with multiple speakers.
Open-source voice AI · Microsoft
VibeVoice aims to turn text into hours-long, multi-speaker conversations
Microsoft's open-source system targets long-form speech rather than short clips. It can produce extended audio with several speakers, a capability the project positions for podcasts, audiobooks and voiceovers.
Hours
Target length of generated audio, far beyond single-passage clips
Multi
Several speakers in one extended exchange
Open
Developers can run the software themselves
From download to long-form audio
The onboarding path is aimed at hands-on developers and production teams.
STEP 1
Download the software
→
STEP 2
Run it in your own setup
→
STEP 3
Generate multi-speaker text-to-speech
→
USE
Podcasts, audiobooks, voiceovers
What the project emphasizes
Long-form generation instead of short voice clips
Consistent multi-speaker exchanges over extended audio
Open-source access, so creators can run it independently
Open questions for adoption
Independent benchmarks and audio quality comparisons
Detailed hardware requirements, including consumer machines
Licensing terms and safeguards against voice impersonation
What to watch
The real test is performance in independent hands and uptake by audio producers.
01 How it performs in independent testing
02 Adoption by podcast producers, publishers and audio developers
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...