AI disclosure
Every clip is synthetic. No real human voice is ever used.
Every clip on Hushlineis synthetic AI-generated audio. No real human voice is ever used, and no recording is ever made. Here's what that means and how we keep it that way.
What Hushline is
A text-to-speech generator with integrity built in.
Generation
A vendor text-to-speech engine renders every clip from typed text. You type a phrase, the engine speaks it in a chosen voice and intensity.
Watermark
Before a clip ships, the pipeline runs a watermark pass that stamps the MP3 with a per-clip identifier. See the next section for how that works.
Age gate
Pages that handle a real-world use case (orders, checkout) are gated behind a one-time age check. The generator itself is open.
Traceability
Every shipped file carries a per-clip id embedded in the audio and in the ID3 metadata. A clip can be traced back to its order row.
How integrity is enforced
Two locks. One promise.
Lock 1 · Watermark
Every shipped MP3 carries a per-clip fingerprint.
After the vendor engine renders the audio, the pipeline runs an ffmpeg pass that appends a half-second inaudible tone tail to the MP3. The tail encodes a one-byte per-clip id and a 16-bit checksum as a 16 kHz on-off-keyed signal at roughly −26 dBFS — below the threshold of consumer speakers, recoverable with a short FFT over the tail.
The same clip id is also written into the MP3's ID3 metadata: the file is tagged with artist = Hushline and a comment field that embeds the id verbatim. The metadata trail survives a re-encode chain that strips the inaudible tail, so the trace is still recoverable from the ID3 alone.
Lock 2 · Age gate
One date-of-birth check. Then it's gone.
Pages that touch an order, a checkout, or any surface that handles a real-world consequence are gated behind /age-gate. The check asks for a date of birth, compares it to an 18+ threshold on this server, and then throws the value away.
What we keep is a single HMAC-signed hl_age_verifiedhttpOnly session cookie — a true/false flag, no name, no email, no birth date. The cookie is private to this tab, lasts only for the session, and never includes any third-party tracking. Closing the browser clears it.
What Hushline is not
The things we deliberately don't do.
- No microphone capture. The browser never records your voice. There is no audio input on any page.
- No user-uploaded audio. Clips come from typed text, not from files you upload.
- No real-person voice cloning. The catalog is a set of licensed synthetic performers. No real person's voice is used, sampled, or imitated.
- No biometric identification. Nothing on the site identifies, classifies, or profiles a person from their voice, face, or any other biometric signal.
That's the disclosure
Synthetic audio. Watermarked at the source. Gated where it matters. If you have questions, the inbox is open.
Back to the generator