Deepfake detection that never leaves your device.

Image, video and voice forensics powered by Transformers.js, NX and pure-DSP heuristics, running 100% in your browser. No accounts, no uploads, no telemetry.

Launch detector How it works

Tools

Three engines, one promise

Image

NX classifier + CLIP zero-shot + JPEG-block & pixel forensics.

Open tool

Video

Frame sampling across the timeline plus the muxed audio track.

Open tool

Voice

Pitch jitter, loudness uniformity, ZCR & vocoder forensics.

Open tool

Private by default

Files never leave your device. Every model runs in WebAssembly.

Forensic ensemble

NX, CLIP zero-shot, FFT, JPEG-block forensics & voice DSP, combined.

Fast & offline

Cached models keep working with no network, with a downloadable report.