Licensing FAQ

How VibeVoice licensing works

VibeVoice core models are released under Apache 2.0. Enterprise deployments can add commercial addenda for private voices, indemnification, and managed compliance. Use this page to map your scenario.

Summary table

Use case License Extras
Research / OSS Apache 2.0 (free) Attribution + GitHub issues
Commercial apps Apache 2.0 + Commercial Addendum Token quotas, SLA, optional white-label
Broadcast / Call centers Enterprise Agreement Indemnity, compliance audits, on-prem nodes

Attribution requirements

All OSS deployments must reference “VibeVoice – Microsoft Open Source Voice Synthesis” in documentation or UI. Include a link to vibevoice.online.

Content policies

Prohibited uses include biometric misuse, impersonation without consent, and content violating regional laws. Enterprise agreements embed auditing APIs.

Private voice cloning

Requires signed consent from the voice talent, stored for 7 years. Upload consent artifacts through the compliance portal or via /v1/consent API.

Common questions

Can I redistribute the model weights?

Yes, under Apache 2.0, provided you keep license headers intact. Commercial add-ons may limit redistribution of fine-tuned voices.

Do I need a license for Spanish or Japanese blueprints?

Blueprint content is covered by the documentation license; however, if you deploy Spanish or Japanese voices at scale, we recommend the localization addendum for regulated industries.