Semantics Deepfake-Detector-Logo
S21 Deepfake Detector​

Assess image and video authenticity when certainty matters.

S21 Deepfake Detector helps investigators assess whether images or individual video frames may be AI-generated, synthetic or manipulated. It analyses visual indicators, presents confidence-based results, and supports informed human judgement - fully offline for sensitive investigative environments.
The Problem

When authenticity is unclear, investigation slows down.

Investigators increasingly encounter images and video where it is not immediately clear whether the material is real, AI-generated or manipulated. That uncertainty can slow down decisions, especially in cases involving safeguarding, sextortion, fraud, online harm or intelligence development.

Without structured indicators, teams are left relying on manual visual judgement alone, making it harder to assess credibility, prioritise next steps and understand the evidential value of uncertain media.
See how it works in 5 minutes

Join a former law enforcement officer for a practical walkthrough of the tool and features designed to support real-world investigations

Fully Offline & Secure

All analysis runs locally. Sensitive imagery and video never leave your environment.

Designed for investigative reality

Built for safeguarding teams, DFUs, and intelligence analysts — not social media moderation.

No specialist infrastructure required

Runs on existing workstations without servers, cloud services, or complex deployment.

We’re creating an open community of collaboration with other industry experts​

S21 VisionX fully supports evidential data exchange with leading digital forensic tools, ensuring seamless integration with your existing workflow. Compatible platforms include X-Ways Forensics, Cellebrite, Magnet Forensics (AXIOM, IEF, Witness), GrayShift, Griffeye, Belkasoft X, Oxygen Forensics, OpenText EnCase, MSAB XRY and XAMN, Exterro, GetData, and many others. With S21 VisionX, you can review, categorise and identify victims efficiently — and easily export results back to your tool of choice for further analysis or reporting.
The Solution

What S21 Deepfake Detector helps investigators do

Assess still images and video frames

Analyse images or selected video frames where authenticity is uncertain.

Identify manipulation indicators

Surface visual artefacts, inconsistencies and patterns associated with synthetic or manipulated media.

Understand confidence clearly

Present confidence-based results, including where the assessment is uncertain or indeterminate.

Support human judgement

Give investigators structured information to assess authenticity without replacing professional review.

Review manipulation indicators faster.

S21 Deepfake Detector gives investigators a structured way to examine uncertain images and selected video frames inside their own environment. Instead of relying on manual visual checks, third-party upload tools or lengthy review, teams can assess indicators associated with synthetic generation or manipulation and receive confidence-based results that support further investigative judgement.

The tool is designed to highlight areas of concern, surface uncertainty clearly and help investigators decide whether material needs further review, escalation or contextual assessment. It does not make final determinations on authenticity - it gives teams clearer information, faster, while keeping sensitive media under investigator control.
How it works

From suspicious media to structured assessment

S21 AutomateX supports teams through the preparation and processing stages of large-scale digital investigations, helping them automate repeatable X-Ways and VisionX workflows, reduce manual handoff, and move evidence into structured review faster.

Without automation, investigators are forced to spend critical time managing process instead of assessing material.

Add the image or frame

Investigators start with a still image or selected video frame where authenticity is in question.

Run local analysis

S21 Deepfake Detector assesses the material using forensic indicators associated with synthetic generation or manipulation.

Review the confidence result

The tool presents results in a clear format, including where confidence is low or uncertainty remains.

Use the assessment in context

Investigators use the result to support safeguarding, intelligence, evidential review or further enquiry.

"The Semantics 21 team clearly put the customer first. It's refreshing to feel that we have a company prepared to listen to our individual needs and create solutions which are fit for purpose."

D/Insp Milligan
Merseyside Police, UK
Features

Authenticity assessment features for investigative review

Media assessment

  • Still image assessment
  • Individual video frame assessment
  • Synthetic media indicators
  • Manipulation indicators

Indicator analysis

  • Visual artefacts
  • Geometric inconsistencies
  • Biological plausibility indicators
  • Micro-textural and statistical signals

Investigator review

  • Confidence-based outputs
  • Low-confidence or indeterminate handling
  • Clear presentation of results
  • Support for human judgement

Secure operation

  • Fully offline analysis
  • Local processing
  • No cloud upload required
  • Suitable for sensitive investigative material
Offline & Secure

Assess sensitive media without cloud upload.

S21 Deepfake Detector runs locally, allowing investigators to assess sensitive imagery and video frames without relying on cloud services or external tools. This supports controlled review in safeguarding, forensic and intelligence environments.
Semantics School Badge Offline Mode
Who it's for

Built for teams handling uncertain media

S21 Deepfake Detector is built for teams that need to assess whether images or video frames are real, synthetic or manipulated.

Safeguarding teams

Assess whether imagery may be synthetic, manipulated or unreliable where safeguarding relevance is unclear.

Digital forensic units

Support structured review of seized media, extracted images or video frames.

Intelligence analysts

Add a consistent authenticity assessment layer when media is being used to inform intelligence decisions.
Operational Capability

Built to assess authenticity, uncertainty and risk

S21 Deepfake Detector supports structured assessment of images and selected video frames where authenticity is unclear. It helps investigators review indicators of synthetic generation or manipulation, understand confidence levels, and keep uncertain media under controlled human assessment.

Image and frame assessment

Assess still images and selected video frames where authenticity is uncertain.

Synthetic media indicators

Surface signs associated with AI-generated, synthetic or manipulated imagery.

Confidence-based results

Present clear results, including where confidence is low or the assessment is indeterminate.

Investigator-led review

Support professional judgement without making automated investigative decisions.
S21 Deepfake detector + S21 VisionX

S21 VisionX, where structured assessment operates at investigative scale

Authenticity assessment is only one part of large-scale investigative review. Investigations involve thousands of files where authenticity may be uncertain.
Semantics Deepfake-Detector-Logo

S21 Deepfake Detector

When authenticity is uncertain, move from suspicion to informed assessment, without relying on external tools or guesswork.

S21 VisionX

The offline investigative platform these tools were built inside.

S21 Deepfake Detector was developed and proven inside S21 VisionX and is now available as a standalone tool due to investigator demand.

Used in investigations where delays and missed connections have consequences.

Designed for secure environments. Built for large-scale investigations. Developed with investigators and forensic practitioners.

Operational impact across real investigations

See how investigative teams are using Deepfake Detector to analyse evidence, identify connections and progress complex cases.

Compare how authenticity is currently assessed

Run your current approach alongside S21 Deepfake Detector on the same material.