Voice-Cloning Consent

Separate and understandable consent

Voice consent will not be buried in general terms or pre-selected. The speaker-facing form must be presented in a language the speaker understands and record its exact version, language, timestamp, speaker and account, purpose, evidence reference, and withdrawal state.

What the speaker confirms

• the recording is their voice, or the submitter has documented authority for the identified speaker;
• Ozelaro and its contracted voice provider may process the recording to create and use a synthetic voice profile;
• the allowed purpose is narration requested by the authorised Ozelaro account;
• the voice cannot be used for impersonation, deception, political persuasion, financial or identity verification, harassment, or an unapproved purpose;
• generated speech can sound like the speaker but can contain errors;
• consent can be withdrawn for future generation through the documented process.

Planned product controls

• a guided 90-120 second sample in the speaker's natural target language;
• speaker identity or authority evidence proportionate to the risk;
• a tenant-scoped voice profile and restricted server-side provider access;
• an audible or text preview and first calibration approval before paid automation;
• versioned consent evidence and an accessible withdrawal route;
• future generation blocked immediately after withdrawal;
• deletion requests propagated to Ozelaro storage and relevant vendors, with confirmation or a lawful-retention explanation.

Withdrawal limits

Withdrawal stops future generation. It cannot automatically remove media already downloaded or published to third-party services, although Ozelaro will support future takedown actions within connected platforms' capabilities. Public figures, third parties, minors, and vulnerable people require a separately reviewed lawful process and are not accepted by default.