Are there any personal voice models that use licensed training data for high-fidelity cloning?

Yes. A few personal voice / voice-cloning systems are built on licensed or otherwise consented training data, though the exact terms vary by provider and product tier.

Examples include:

• ElevenLabs — offers voice cloning and has emphasized consent/licensing workflows for many of its voices and datasets.
• Resemble AI — supports custom voice cloning and positions itself around consent-based voice creation.
• Play.ht — provides voice cloning and enterprise offerings; some voices/data are licensed or consented depending on the setup.
• WellSaid Labs — focuses on licensed, studio-recorded voices rather than user-supplied cloning in the same way.
• Descript Overdub — personal voice creation from your own recordings, which is consent-based rather than third-party licensed.

A key distinction:

• Licensed training data means the model was trained on data the vendor has rights to use.
• Personal voice cloning often means the model is trained on your own voice recordings, which is consented but not usually called “licensed” in the same sense.
• For high-fidelity cloning, most reputable services require explicit consent and have guardrails to prevent impersonation.

If you want, I can also give you:

1. a comparison table of the main providers,
2. which ones are best for realistic personal cloning, or
3. what to look for in their licensing/consent terms.