Someone asks: “Is this real?”
Without provenance, your word is all you have.
Detectors answer after the clip is already spreading, and only with a probability. SIGNED SOUND starts earlier. Before you publish, a provenance mark goes into your official audio and its acoustic features are recorded. When a clip surfaces later, you check it against what you registered, and show, as a verification result, whether it came from the registered source and whether it differs from the registered original.
We check one thing: whether the audio came from the source that registered it, and whether it differs from the registered original (authenticity). We do not judge whether what is being said is true, or whether the speaker is who they appear to be.
Shachihata embeds the mark. You run the checks yourself, in the system we provide.
The pilot costs nothing. Production is quoted per project. You don’t need a budget, or even an approved project, to talk to us, and we won’t chase you.
The service is offered only in Japan, and is provided only for Japan-based projects. We are happy to respond to inquiries in English.
*A comparison against registered data. Not a guarantee of overall authenticity.
Six places this problem shows up
SIGNED SOUND does one thing: it lets the source of a recording show where that recording came from. Whatever the sector, the underlying need is the same — keep what you released in a form you can check against later. Any of these sound familiar?
Audio of an evacuation order, an election-commission announcement, or a mayor’s press conference gets shared around, and someone asks whether it is genuine.
Details for local governments (Japanese) →A clip of something you broadcast spreads with the caption “did they really say that?”
You have wondered what happens the day someone clones your chief executive’s voice.
You suspect audio you published is being reused somewhere you never authorized.
A recording arrives, supposedly from a colleague. You have no way to check.
Your radio segment or podcast gets clipped, and afterwards you cannot tell what was cut.
All six have the same starting point.
Tell us yoursYou can’t detect your way out of this.
Detection has its place. But when the recording is a public safety notice or an earnings call, “our tool says it is probably fake” is not an answer anyone accepts. And provenance only exists if it was recorded before the audio left your hands. Once the clip is spreading, that window has closed. The question you get asked afterwards is rarely why you had no countermeasure. It is why you didn’t take the one that existed.
- Chases each new generation method as it appears
- Calls a cut, reordered, or spliced clip genuine, because nothing was synthesized
- Only responds once the clip is already spreading
When the stakes are high, “probably fake” settles nothing.
- Your official audio is marked before you publish, and its acoustic features are recorded
- Clips and copies can often still be matched back to the registered original
- You can show whether audio came from the registered source, and whether it differs from what you registered
The provenance exists before the incident does.
None of this argues against detection. Detection is how you assess someone else’s content once it is out in the wild. Provenance is how you mark your own before it leaves. Most organizations will want both.
For reference: in an analysis of roughly 126,000 news cascades on Twitter, true stories took about six times as long as false ones to reach 1,500 people (Vosoughi et al., Science, 2018). *A study of news diffusion in general, not of audio specifically.
Register it before you need it.
A digital watermark, designed to be hard to notice in normal playback, goes into your official audio, and its acoustic features are recorded. If a clip surfaces later, you check it against the registration and show whether it came from the registered source and whether it differs from what you registered. One step taken early is what keeps you from having to react from behind.
Mark it before you publish
Before release, the mark goes into the original audio and its acoustic features are recorded. Perceptible impact on sound quality is kept low.
Check the clip against it
Circulating audio is compared with the registered original. Copies that lost the mark can often still be matched by their acoustic features.
See where it came from
You get a verification result: the registered source, and whether the audio differs from what was registered.
A watermark and a fingerprint, both built for audio
Matching works from the waveform itself, not from metadata that any re-upload can strip. *Re-encoding and processing by distribution platforms can defeat matching, depending on the type and degree. Matching under every transformation is not guaranteed.
A reference point that predates distribution
The moment you register is the reference point. Because it is recorded before the audio goes out, it becomes your own anchor for checking, later, when and where a clip came from.
No special equipment, no new recording kit
It does not matter what you recorded on, or where you publish. The mark can be added once the recording exists.
“It still can’t tell me whether what is being said is true.” Correct. SIGNED SOUND answers one question: did this audio come from the source that registered it, and does it differ from what was registered (provenance and authenticity)? It does not judge whether a statement is factual, or whether the speaker is who they appear to be, and it does not stop anyone from making a fake in the first place. We would rather be clear about the limit. The one question it does answer, you can answer yourself, at the moment you need to.
What changes when the clip starts spreading
What you get is a position you prepared earlier: the origin of your official audio, and any differences from what you registered, can be shown, grounded in a verification result. It does not take disinformation down, and it does not judge whether content is true.
| Situation | Before | After |
|---|---|---|
| Someone asks: “Is this real?” | All you can do is point at the official page | You show the registered source and any differences, grounded in a verification result |
| Your audio spreads on social media | You cannot be sure whether it is yours or an edit | You can often match it against the registered original and see whether anything differs (heavy conversion or processing can defeat matching) |
| What it covers | – | Provenance and authenticity, and nothing else. No takedowns, no judgments about truth |
Can your audience check the source?
Your listeners, whether residents, viewers, or business partners, get a starting point for checking, through the trust mark, whether audio came from a registered source and whether it differs from the registered original. It is not an instant verdict on the device. And “unregistered” does not mean “fake”: it means nobody registered it.
Under the defined verification conditions, no differences from the registered data were found.
The audio differs from the registered data. What caused the difference, and whether anything was tampered with, needs a separate look.
Nobody registered this content.
*Where the mark appears, and what the check screen looks like, depend on the distribution channel and the implementation (illustrative screen). Listeners are expected to be able to check without a dedicated app.
Who is behind this
Technology partner
The acoustic identification — embedding the mark, matching the features — comes from Evixar Inc., a specialist in acoustic watermarking. Evixar’s EAW and EFP are listed on the C2PA Soft-Binding Algorithm List (registered IDs: com.evixar.eaw.1 / com.evixar.efp.type3.1).*This means the algorithms appear on that list. It does not mean this service has obtained C2PA conformance certification, or that Evixar or Shachihata are members of C2PA/CAI.
Field trials
Under a program of Japan’s Ministry of Internal Affairs and Communications, we ran a proof-of-concept with Kyoto Prefecture on the authenticity of official audio. We have no production reference cases yet. We would rather tell you that up front.
Operator
Shachihata is the company behind the name seals used across Japan. For a century our business has been marks that carry trust. Here we supply third-party reference information for checking provenance.*Reference information to support verification of content provenance and authenticity. It is not a judgment on the truth of what is said, and it does not create or confer any statutory presumption — in particular, not the presumption of authentic establishment of an electronic or magnetic record under Japan’s Act on Electronic Signatures and Certification Business (Act No. 102 of 2000).
Start with a conversation, not a contract.
The pilot costs nothing. Production is quoted per project. We look at the audio you actually publish, and propose something that adds as little work as possible.
Frequently asked questions
Does it also tell me whether what is being said is true?
Is it admissible as evidence in court? Does it have legal effect?
How is this different from a deepfake detector?
What is the “free pilot”?
Where is the audio data stored?
Do we have to change how we publish today?
Record the provenance before you need it.
Ask for a conversation, request the materials, or set up a pilot. The pilot costs nothing; production is quoted per project. You don’t need a budget or an approved project to talk to us, and we won’t chase you. The service is offered only in Japan, and is provided only for Japan-based projects. We are happy to respond to inquiries in English.
(1) This service verifies only integrity (sameness with the registered data) and provenance, from the point of registration onward. It does not prove that what is said is true, and it does not prove who the speaker is. The source that registered the audio is not necessarily the speaker.
(2) Use of this service does not guarantee admissibility as evidence in court or any specific legal effect.
(3) Authenticity (sameness with the registered data and the existence of provenance) is distinct from whether what is being said is true. We do not judge truth.