Voice cloning, consent, and why VibeSing is built differently

Voice cloning technology is powerful enough to be genuinely unsettling when misused. We built VibeSing with that in mind — every design decision about voice data starts from the question: would this person consent to this?

The consent-first rule

The only voice you can clone on VibeSing is your own — recorded live, in the moment, from your own device. There is no upload-a-random-audio-file path. This is a deliberate design choice, not a technical limitation. It makes impersonation meaningfully harder.

What we store and why

Your voice samples are uploaded to a private Vercel Blob store under your account. They are used to train your voice model on Replicate's infrastructure. After training, the raw samples remain linked to your account. You can delete your data at any time from the User Center.

• Raw samples are stored per-account — not shared across users
• Trained voice models are private to your account
• Generated clips can be shared publicly only when you choose to share them
• We do not use your voice data to train general models

Sharing and public visibility

When you share a clip, the audio and a preview image become publicly accessible via the share link. The share can include your display name if you choose. You control the headline, dedication, and whether the share link is visible at all. Share pages are currently not indexed by search engines.

Why this matters for AI music

The broader AI music space has a consent problem. Tools that let anyone clone anyone else's voice from a public audio clip cause real harm — to artists, to public figures, to ordinary people. We think the right answer is not to make cloning harder technically, but to make consent structural. If the only voice you can use is one you record yourself, the harm surface collapses.