VoiceSign is a small toolkit, with source code on GitHub and a Python package on PyPI, that lets producers cryptographically sign voice recordings and lets recipients verify them.
How it works
A producer signs an audio file with their private key and an identity string. The signature is embedded as an inaudible cryptographic watermark in the audio bytes themselves -- the file size and audible quality are unchanged. Anyone receiving the file can run the watermark recovery and check the signature against the producer's public key.
The verifier ecosystem
To make verification one click rather than a developer task, ORAVYS hosts a public registry at oravys.com/s/<id>. Producers get a shareable URL back. For anyone opening it, the page shows only the signature status, the signing date, the asset id and the SHA-256 fingerprint of the signed file, with a local checker that hashes a file in the browser without uploading it. The audio itself and the signer's self-declared identity are visible only to the signer's own account; a DNS-verified domain is shown because the signer published it.
Honest limits
Identity claims are user-asserted by default. A signer who writes 'Elon Musk' as their identity is not magically Elon Musk. To get a verified identity badge, signers can publish a DNS TXT record on a domain they control. Without that binding, every identity is displayed as 'self-claimed, not independently verified'.