●Legal · Litigation & investigations

Screen digital evidence before it enters proceedings.

Know which exhibits warrant forensic review. Witness returns a likelihood, a per-model breakdown and the signals it found, a strong first signal alongside expert analysis.

Multi-model detection<3s per imageZero retention
exhibit_14_deposition.mp4scan complete
Submitted evidence
blend boundary
geometry drift
frames 212–244
AI-generated
0.89LIKELIHOOD
Refer to forensic expert
Face-swap boundary
91
Temporal consistency
84
Lip–audio sync
66
C2PA / EXIF
32
<3sMedian result time
4Independent signal families
0Files retained
C2PAProvenance read

●Platform

Detect it. Explain it. Act on it.

Every scan returns the same three things: a result, the signals that produced it, and a decision you can route on.

exhibit_14_deposition.mp4intake · 10:42:08
AI-generated
exhibit_15_deposition.mp4intake · 10:42:11
Real
exhibit_16_deposition.mp4intake · 10:42:15
AI-generated

01

Detect

Every file is scanned by independent models for images, video and faces, in the web tool, the extension or by API.

Face-swap boundary
91
Temporal consistency
84
Lip–audio sync
66
C2PA / EXIF
32

3 / 4 signals corroborateCalibrated likelihood

02

Explain

See which signals fired and how strongly. No single model decides; every result comes with its reasons.

Clear
Refer
Review

Rule: likelihood ≥ 0.85 → Refer<3s

03

Route

Branch your workflow on the result. Clean items pass, flags go to review with a full audit trail.

●The problem

Digital evidence is getting harder to trust.

Courts are beginning to address AI-generated evidence, with proposed rules such as FRE 707 and amendments to FRE 901(c). Counsel need a way to evaluate media before relying on it or presenting it.

96%Of people cannot reliably identify AI-generated faces
400%Increase in deepfake incidents since 2023
901(c)Proposed FRE amendment for digital evidence
32US states with deepfake-related legislation introduced

●What we catch

Five layers of analysis.

Flagged at the point of submission, before it reaches a reviewer.

01

AI-generated images

Pixel patterns, frequency artefacts and fingerprints consistent with diffusion models and GANs.

02

Face swaps in video

Frame-by-frame detection of replacement artefacts, blending boundaries and temporal drift.

03

Manipulated video

Splicing, frame insertion and temporal manipulation; inter-frame compression consistency.

04

C2PA provenance

Verification of signed Content Credentials, capture device to current file, when present.

05

EXIF & metadata

Mismatched timestamps, missing camera data, generator signatures and signs of stripping.

The pipeline

A detector for each kind of media.

Image, video and voice each have their own detector. Their findings come back as one clear answer you can act on.

Image

Diffusion, GAN and face-swap images from Midjourney, DALL·E, Stable Diffusion, Flux and Firefly.

Live

Video

Face swaps across 32 frames and fully generated video from Runway, Kling, Veo and Pika.

Live

Voice

Cloned and synthesised speech in calls, notes and recordings.

Live

What you get

Each file is marked Clear, Refer or Review, with the likelihood, signals and audit ID behind it.

One answer

●Important limitations

What Witness is, and is not.

Witness is a strong first signal. It tells you where to look harder.

Witness is

  • A screening tool that identifies statistical patterns consistent with AI generation
  • A likelihood with a per-model breakdown, never a certainty
  • A way to decide which evidence to send to a qualified expert

Witness is not

  • A Daubert-standard report
  • A claim of court admissibility
  • A replacement for chain-of-custody procedure or forensic examination

●Practice areas

Where it applies.

Screen exhibits before filing and flag manipulated media early in discovery, before it becomes a surprise at trial.

Evaluate photos, screenshots and video in custody disputes, protective orders and divorce proceedings.

Identify prosecution exhibits that warrant further forensic examination.

Triage whistleblower evidence, complaint documentation and digital communications at volume.

●How it works

From submission to decision.

01

Upload evidence

Images or video through the web tool. No installation required.

02

Multi-model analysis

Independent detectors check fingerprints, frequency, face-swap boundaries and metadata.

03

Review the breakdown

A 0–100 likelihood and the signals behind it, per model.

04

Document & refer

Record screening as due diligence and refer flagged media to your expert.

●Integration

Three ways in.
Start with the web app.

Scan in the web app or Chrome extension today. The REST API and MCP server for your AI agents are coming soon. Same detectors, same result, same audit trail.

Web app

Available

Upload in the web app. Images, video and voice, up to 50 MB. Full results in seconds.

witness.vision/scan
Page media● Scan
exhibit_14_deposition.mp4 · AI-generated · 0.89

Chrome extension

Available

Scan any image or video from inside the tools your team already uses. No download, no upload.

portal.example.com
Page media● Scan
exhibit_14_deposition.mp4 · AI-generated · 0.89

API

Soon

Single scans, batch jobs and webhooks. Structured JSON with result, likelihood, signals and audit ID.

POST /v1/detect
curl -X POST https://witness.vision/api/v1/detect \
  -H "Authorization: Bearer $WITNESS_KEY" \
  -H "Content-Type: application/json" \
  -d '{"media_url": "https://…/exhibit_14_deposition.mp4"}'

→ 200 OK · 412ms
result: ai_generated · 0.89

●Why Witness

Built to be checked.

We say what the system does, what it returns and where its limits are.

Specialized models

Detection built per signal

Specialized models per signal, so you get more coverage and fewer false positives than one classifier.

Explainable

Reasons, not just a score

Every result shows the signals behind it and where in the file they fired.

Zero retention

Analysed, then gone

Media is never stored, shared or used for training.

C2PA

Provenance, when it exists

Reads Content Credentials from capture device to current file.

●Pricing

Priced around your volume.

No minimums and no long-term contracts. Tell us your volume and we will quote within one business day.

Free

$0

  • 20 image scans a month
  • 2 min video/voice total
  • Full heatmaps
  • Generator attribution
  • EXIF/C2PA/SynthID
  • Web and Chrome extension
Create free account

●Pro

$15 / month

  • 1,000 image scans a month
  • 15 min video + 10 min voice
  • Everything in Free
  • Faces and face-swap specialist
  • Scan by URL: paste any link
  • Email support
Get started →

Enterprise

Custom

  • Volume pricing
  • Everything in Pro
  • SSO and audit logs
  • Dedicated support
  • Custom SLA
  • On-premise option
Talk to sales

●FAQ

Common questions.

Something else? Talk to our team

Can AI-generated evidence be detected?

AI-generated media leaves statistical patterns detection models can identify. Detection is probabilistic, so results should inform further investigation, not replace expert testimony.

Is Witness admissible in court?

Witness is a screening tool, not a certified forensic instrument, and does not currently produce Daubert-standard reports.

What formats are supported?

Images (JPEG, PNG, WebP) and video (MP4, WebM), plus C2PA and EXIF inspection.

Can opposing counsel challenge results?

Yes. That is why we recommend pairing Witness screening with expert forensic analysis for any evidence presented in proceedings.