PRVNZ Fingerprint
Who owns this recording?
Drop any audio. PRVNZ finds who registered it, the file itself or the same recording in another file, cut, re-encoded or turned down, and what they say about AI training. Every answer is signed by PRVNZ and checked in your browser.
How it works
- The file. Its SHA-256, and for lossless audio the hash of the samples themselves, which is the same whether the audio is saved as 16-bit, 24-bit or float.
- The sound. A fingerprint of about eight values a second (Chromaprint's algorithm), which survives MP3 and AAC encoding, volume changes, noise, filtering and resampling, and finds a clip inside the full track, with where it sits. It carries an ISCC Audio-Code (ISO 24138) too.
- The claim. The owner registers it with their PRVNZ account or, stronger, with the PRVNZ Fingerprint app, where their machine's hardware key holds the statement. The earliest registration is the owner of record; later ones are its history.
- The limit. A transposed copy (pitch-shifted by a semitone or more) is not reliably found yet. The fingerprint shows the recording is the same; who made it is what the owner says, signed and dated.
For AI developers: check before you train
A rights holder's reservation is bound to the sound, not to a URL, so it survives a re-upload. Send a fingerprint and read the owner of record's word: train_ai is n (reserved) or y (allowed), in the IETF AIPREF vocabulary, with W3C TDMRep's tdm_reservation and an optional licensing policy. Each answer carries PRVNZ's Ed25519 signature over the statement.
POST https://license.prvnz.com/match
{ "fingerprint": "<base64 of little-endian u32 values>" }
→ { "iscc_audio": "ISCC:EIA…",
"matches": [ { "statement": "{… \"reservation\": {\"train_ai\":\"n\",
\"tdm_reservation\":1,\"policy\":…} …}",
"signature": "…", "public_key": "lmRqUGTQ…",
"match": { "at_seconds": 72.4, "overlap_seconds": 9.8,
"mean_differing_bits": 1.1 } } ] }
GET https://license.prvnz.com/verify/<sha256> by file or audio hash
GET https://license.prvnz.com/iscc/<code> by ISCC Audio-CodeThe fingerprint is Chromaprint's default algorithm on mono audio at 11025 Hz; the PRVNZ SDK and the WebAssembly module this page uses compute it identically.