Voiceprint
Also known as: speaker embedding, voice profile
A voiceprint is a compact set of numbers that captures what makes one speaker sound like themselves, separate from the words they say. A cloning model uses it to render new text in that voice. It is closer to a fingerprint than to a recording.
Why it matters
The voiceprint is what makes a clone reusable. Because it separates who is speaking from what is said, you can build it once and then read any script in that voice without going back to the original recordings. It is the asset, and the recordings are just how you make it.
It also carries a privacy weight. A voiceprint identifies a person, so where it is stored and who can use it is a real question, not a technical footnote.
In practice
Treat a voiceprint like a credential: keep it somewhere you control, and be deliberate about services that upload or store it. The cleaner and more representative the source audio, the better the print captures the full range of your voice.
How Vocast handles this
Vocast builds a voice profile from your recordings and keeps it on your machine. It is the reusable asset you version and roll back, and it never leaves the device.