Every utterance leaves a receipt. Every voice-print consent gets signed.
A voice merchant on the Visa CLI mints one hive-vcr-1 receipt per synthesis. The voice-print consent, the source-script hash, and the likeness rights all get co-signed before the audio comes back.
These are the fields the operator co-signs before the call returns.
Every field below gets co-signed by the operator with ML-DSA-65 and countersigned by Hive. If a field is missing, that's not a default value. It's a verification failure. The merchant can't return the call's result without the receipt, and the receipt can't exist without these fields.
Voice / TTS · attestation fields
Three places this business needs a receipt right now.
Voice-clone compliance
When a brand licenses a celebrity's voice, the per-utterance receipt is the proof the talent's agency can check every quarter, without having to trust the brand or the vendor doing the synthesis.
IVR and customer service
A regulated phone system can prove exactly what was said and when, in a form regulators accept without having to subpoena the vendor.
Accessibility narration
Public-sector accessibility narration carries its consent and rights information right in the receipt. The receipt log becomes the procurement file.
One endpoint, one envelope, and no dependency on Hive in your inference path.
A voice or TTS operator can adopt hive-vcr-1 by adding the receipt envelope to the response and publishing a verification key. Hive's countersignature rides alongside, it's not on the critical path. So if Hive ever disappears, the operator's receipts still check out against the operator's own key.