Home // FORENSIC UNITS // VIDEO, VOICE AND IMAGE

Audiovisual Forensics: Video and Image

Video shows what happened, or what someone wanted it to look like. The examination, aligned with the international SWGDE and ENFSI practices, answers which of the two: global analyses of the file, local analyses of the content and chain of custody from the recorder to the expert report.

What audiovisual forensic examination answers

Nine examinations on video and image

Every forensic question has an examination of its own: is the file authentic? Was it edited? Who appears in it? What is the real measurement? The expert report declares the method and the limit of each answer.

01

Video authenticity and integrity

Is the file what it is said to be? Analysis of the container structure, metadata read with multiple tools (SWGDE 23-V-001 practice), encoding and recompression history, coherence of shadows, lighting and edges, temporal continuity.

02

Detection of editing and cuts

Jumps and gaps in the frame sequence, GOP and compression discontinuities, duplicated or suppressed frames, inconsistencies between the audio and video tracks.

03

Video deepfakes and synthetic media

Facial blending artefacts, inconsistencies in blinking and physiology, mismatch between lips and speech (phoneme against viseme), spatial, temporal and frequency anomalies, combined with the cloned-voice examination from forensic phonetics: the complete detection pair.

04

Image analysis (photography)

EXIF and metadata, traces of tampering (cloning, splicing, error level analysis and the like), coherence of the compression history and linking of the image to the equipment that produced it.

05

Photogrammetry and facial and body comparison

Real measurements extracted from the image: height, distances, vehicle speed. In prosopography, anthropometric points of the face, individualising marks (tattoos, scars) and gait pattern, with a probabilistic conclusion, never "absolute recognition" from low-resolution footage.

06

Enhancement and clarification

Sharpness and contrast enhancement, stabilisation, licence plate reading and forensic magnification. Processing reveals what is recorded; it never creates content: the same ethics as audio enhancement.

07

Audio-video synchronisation

Does the soundtrack belong to this video? Examination of mismatch, suppression or replacement of the track, with a natural bridge to phonetic examination when a voice is in dispute.

08

CCTV and DVR analysis

Proprietary extraction from the DVR (the menu export recompresses), verification of timestamps and dating, frame-by-frame examination, ordering of events and chain of custody from the recorder to the expert report.

09

Complementary examinations

Technical description of audiovisual content (forensic narrative of scene and events), reconstruction and Visual Law for presentation in court, forensic redaction of video (masking of faces and licence plates for court use and LGPD (Brazil's data protection act), compliance), verification of date and time from traces in the file, rebuttal of another expert's audiovisual report.

From the frame to the pixel: the examination of video and image
How the examination is carried out

Global and local analyses, converging methods

The examination follows SWGDE best practices (18-V-001 for video analysis, 23-V-001 for authentication, 16-I-002 for images): a defined set of analyses on the file and on the content, with multiple converging methods and bias mitigation.

Global analyses: the file
  • Container structure and format signature
  • Metadata examined with multiple tools
  • Hashing and documented chain of custody
  • Encoding and recompression history
Local analyses: the content
  • Frame-by-frame examination of the sequence
  • Coherence of the scene: shadows, lighting, perspective
  • Physics of light and geometry of objects
  • Temporal and motion continuity

The chain of custody runs from end to end: acquisition of the original (ideally from the device or the DVR), hashing, working copy and a record of all processing, ensuring reproducibility. Conclusions are probabilistic here too, especially in facial comparison: the expert report declares the degree of support given by the evidence, not certainties of convenience. Quality control accompanies the whole workflow: double-checking of findings, a record of every processing step and declared mitigation of contextual bias.

What sets it apart

Forensic collection at the source: audiovisual and computer forensics under the same signature

Extraction of the raw file straight from the mobile phone or the DVR, in the laboratory or in the field, with hashing, binding to the equipment and documented chain of custody, is part of the same forensic work. The examination begins with the original, not with the recompressed copy that circulated through messaging apps.

Methodological grounding: SWGDE 18-V-001 · SWGDE 23-V-001 · SWGDE 16-I-002 · ENFSI practices for forensic image and video analysis.

Declared limits

A natural artefact is not editing, and magnification does not create pixels

Freezing, dropped frames and pixelation from re-encoding (VoIP, streaming, messaging apps) are normal compression behaviour and do not prove tampering. The examination distinguishes the natural artefact from the trace of editing, and that distinction brings down alarmist reports.

An alarmist report confuses artefact with tampering

A video that froze or pixelated is not, in itself, proof of editing. A technical rebuttal shows when the supposed trace of manipulation is merely the codec doing its job.

Material required

What the examination needs in order to begin

OriginalThe raw file from the device or the DVR, not the video downloaded from WhatsApp, Instagram or YouTube: recompression erases metadata and traces
CCTVExtraction from the DVR itself, if possible with the proprietary player and the equipment settings; the menu export recompresses
DeliveryOnly via cloud services (Drive, WeTransfer or a secure link), never through a messaging app; the size of the video does not justify sending it compressed
ContextSource equipment, approximate date and time, what is being questioned (editing? identity? measurement?) and the existence of other copies
Facial comparisonStandard photographs of the person being compared, at an angle and in conditions close to those of the questioned image
On-site collectionWhen necessary, forensic collection takes place on site: extraction from the mobile phone or the DVR with hashing and chain of custody, computer forensics under the same signature
Who it is for

Where audiovisual forensic examination decides cases

Criminal

CCTV, authenticity and measurements

CCTV of crimes: who appears, what happened and when; authenticity of incriminating video, photogrammetry of height and speed, licence plate reading.

Employment

Accidents and dismissal for cause

Video of a workplace accident (dynamics and speed), CCTV in dismissal for cause, authenticity of a recording used against an employee or employer.

Family and civil

Custody, defamation and road traffic

Videos in custody disputes, verification of tampering and deepfakes in defamation, road accident footage.

Corporate

Fraud, reputation and compliance

Internal fraud captured on CCTV, deepfakes in video calls (CEO fraud), verification of media in a reputational crisis, evidence for compliance purposes.

Frequently asked questions in this area
Can it be proved that a video was edited?
Yes: editing leaves traces in the file structure, in the compression and in the continuity of the frames. The examination combines global analyses of the file with local analyses of the content, through multiple converging methods.
And that it was NOT edited?
The documented absence of traces, verified by multiple methods, supports the integrity of the recording. It is an examination in as much demand as the first, and just as decisive.
Is a video downloaded from WhatsApp usable?
With limitations: recompression erases metadata and traces. The ideal is the original from the device, and extraction with chain of custody is part of the service: audiovisual and computer forensics under the same signature.
Can someone be identified in security camera footage?
It depends on the quality: the examination compares anthropometric points and individualising marks and concludes in probabilistic terms. Low resolution may lead to an inconclusive result, and the expert report says so frankly.
Can a dark or blurred video be improved?
Up to the limit of what the sensor recorded. Processing enhances what is there (sharpness, contrast, stabilisation); it does not create what was never captured.
How is a video deepfake detected?
Synthesis artefacts on the face, physiological inconsistencies, mismatch between lips and speech, compression anomalies, together with the digital traces in the file and, where there is a voice, the phonetic examination.
Is it possible to measure height or speed from a video?
Yes, through photogrammetry: with scene references, the examination extracts real measurements (height, distances, speed) with a margin of error declared in the report.
Completed cases
PUBLIC CASE

Expert interviews on election deepfakes for Domingo Espetacular and Band News

Participation as a specialist in news features on the use of AI to manipulate video and audio in electoral contexts.

Complete multimedia evidence

The image is one part. The rest of the evidence can be examined too.

Send your video for a feasibility assessment

The initial screening is free of charge and states frankly what the material allows: origin, resolution, compression and the real chances of reaching a conclusion. Answer within 48 business hours.

Request a free assessment
Representative cases

The level of work you are engaging

Before deciding, it is worth seeing what has already come through this laboratory: cases described without identifying the parties, in the format of challenge, method and result.

See all representative cases →

The technical evidence your case requires. The authority courts respect.

Initial feasibility consultation at no cost. Reply within 24h on business days.
Request an Examination