DetectifAI
Voice identity  //  2026 Live demo running

Every voice is now forgeable. Prove which ones aren't.

DetectifAI builds foundational models to detect audio deepfakes — and the trust layer that puts them to work. Verify the authenticity of voice notes, calls and recordings, and prove that a live speaker is who they claim to be.

Live pipeline Verified
Liveness spoke today's sentence
Speaker identity matches the voiceprint
Audio authenticity no synthesis found
Gates
3

Sequential — all must pass

Identity
1:1

Claim checked against a voiceprint

Trust layer
Layered

Falls back on an uncertain score

Languages
EN + HI

Challenge and transcription

01 / The problem

Voice used to be proof. Thirty seconds of audio is now enough to take that away.

Every call centre, every KYC flow, every "it's me, approve the transfer" voice note rests on an assumption that stopped being true: that a voice is hard to fake. Open cloning models reproduce timbre, cadence and accent from a short sample, and the output survives compression, phone codecs and background noise well enough to pass a human listener.

The defensive side has not kept up. Most teams either bolt on a single deepfake classifier and treat its number as truth, or they check identity and never check authenticity at all — which means a perfect clone of an enrolled user walks straight through the front door.

Cloning is commoditised

A convincing clone costs nothing and takes under a minute. The attacker does not need to be sophisticated.

Replay is the easy attack

Before anyone reaches for a generator, they simply re-submit audio the real user already produced.

One model is one blind spot

Every detector fails on some generator, some codec, some room. A single classifier is a single point of failure.

02 / What we build

We don't ship a detector. We ship the layer that decides.

Detection is not a model, it's a decision — and a decision needs a system around it. The trust layer scores every sample, recognises when that score isn't confident enough to act on, and falls through to another detection layer rather than guessing. One layer answers, the answer is recorded, and the caller gets a decision instead of a raw number.

One verdict, always attributable

A single layer's score decides the outcome, and the audit log records which one. No silent averaging, no blended number hiding a disagreement.

Measured, not assumed

We characterise where each layer degrades — codec, compression, room, generator — and route around it, rather than trusting one model everywhere.

Swappable by design

Layers are configuration, not architecture. Adding, reordering or removing a detector is an environment change, never a rewrite.

Fails the way you choose

Fail-open with a logged warning keeps a call centre moving; fail-closed makes authenticity a hard gate. Same code, one setting.

A verdict, in full
POST /verify/single { "success": true, "reason": "verification_success", "sentence_score": 96, // gate 1 — liveness "similarity": 0.782, // gate 2 — speaker "deepfake_score": 0.041, // gate 3 — authenticity "deepfake_status": "AUTHENTIC" } // Which layer produced the verdict is recorded in the server-side // audit log, never in the response — the caller gets a score and a // decision, not the machinery behind it.
03 / How it works

Three gates, in order. All of them must pass.

Cheapest rejects first ↓

Gates run sequentially and short-circuit: a failure at gate one means the speaker model never runs, and a speaker mismatch means we never spend an authenticity call.

01
Challenge sentence — liveness and anti-replay
The server picks a random sentence, signs it into an HMAC token bound to the user, the purpose and the language, and never trusts the client to say which sentence was expected. The token is single-use and expires in 300 seconds. Speech-to-text transcribes the recording, and the transcript has to match the sentence inside the token.
english + hindi · TTL 300s · single-use · fuzzy transcript match
REQUIRED
02
Speaker identity — the voiceprint
Audio is normalised to 16 kHz mono and reduced to a compact voiceprint that encodes how a person's vocal tract sounds rather than what they said. Cosine similarity against the enrolled template has to clear the bar. Enrollment builds that template from one continuous passage, or from five shorter samples merged together.
16 kHz mono · cosine similarity · CPU inference
REQUIRED
03
Audio authenticity — the trust layer
Two detection layers are attached today. The first scores how synthetic the sample is; if that score comes back as uncertain behaviour, the sample falls through to the second layer, which acts as the fallback and produces the verdict. Audio judged synthetic is rejected even when the voice matches perfectly — this is the gate that stops a flawless clone of a legitimately enrolled user.
two layers · uncertain score falls through · verdict recorded
REQUIRED
Enrollment

One recording of a ~15 second passage, or five short sentences merged into a single template. Each sample passes the same liveness gate before it is allowed to shape the voiceprint, so nobody enrolls a clone.

Attack simulation

The console can submit a pre-recorded or synthetic file through the exact pipeline a live microphone uses. It's the fastest way to watch gates one and three do their job — and it's built into the demo.

Try it yourself

Enroll your voice. Then try to break it.

The console runs the real pipeline — the same gates and the same trust layer described on this page. Record a passage, verify against it, then inject a file and watch the layer reject it.