← Back to overview

For an audio platform · trust & safety

How do you prove a recording is real — and catch the moment it was altered?

Provenance that travels with the file — the seal is part of the audio.

Today audio authenticity relies on metadata that's easily stripped, or detectors that only guess — nothing travels with the file itself.

Why now cheap voice cloning makes 'is this real?' a daily question, just as provenance standards are taking shape.

Use cases Journalism, legal & compliance, insurance, courts, media platforms.

Interactive demo ↗

The question

A recording can be trimmed, spliced or regenerated, and the file looks the same. “Trust me” isn’t evidence.

The approach

Slice the audio into a chain and seal each slice. The seal rides inside the file — nothing to look up, nothing to phone home. Re-cut any slice and the chain says so.

The feature

A live read-out — chain integrity, slices, altered slices, sealed-at — over a signal field of the whole recording. Tamper with one slice and integrity drops on the spot.

Function

The seal read-out

attest — chain integrity
Loading…

Chain integrity and slices in the chain, read straight from the file — not asserted, measured.

Function

Re-cut a slice, the seal breaks

chain integrity 94% altered slice · #4

Alter one slice and the chain no longer matches its seal — the tampered slice is flagged and integrity drops.

Newsletter

Notes from building at the human–AI interface.

Every two weeks: new builds, notes, explorations — the interaction patterns that are working for agentic AI, and short field notes you can put to use.

Subscribe