Content Credentials & C2PA
Signatures, editing actions and available provenance records are examined separately from content-based analysis.
Limit / Missing credentials are not proof of manipulation.
FORENLYX / Evidence system 0.5Image analysis online
Digital media forensics
See whatchanged.
Forenlyx examines images and videos for generative editing, synthetic content and provenance signals — with traceable evidence instead of a single score.
ModuleImage / Beta
AccessTrial + plans
OutputEvidence report
What makes the product different
01Evidence, not an oracle
02Uncertainty made visible
03Images, video and provenance
The public example analysis shows the workflow. Personal images are processed after sign-in with an active trial or plan.
Forensic workspace
Example mode · simulated findings
portrait_0426.jpg/00:02.41
EXIF checkedC2PA absent
Localization 02
Local editing in the face and hair region
Why Forenlyx
01A score is not evidence.
A single AI score conceals which signals shaped the assessment and which counter-signals may exist.
Forenlyx is designed to separate observations, provenance records, model assessments and uncertainty. The result becomes a traceable basis for decision-making rather than a black-box verdict.
Sample report
Observation.
Interpretation.
Limitation.
The report deliberately separates measured signals from their interpretation. It shows where evidence is strong, where counter-evidence exists and when manual review remains appropriate.
CASE / 0426-AReview
Overall assessment
Signals of local generative editing.
The strongest deviations appear in the face and hair region. Consistent metadata acts as counter-evidence. Provenance records are absent.
Manual review recommended
01Content CredentialsNot presentNeutralNo provenance record; no conclusion about manipulation.
02Local image signalsElevated deviationSignalLocal texture and edge patterns differ from their surroundings.
03MetadataConsistentCounter-evidenceTimestamps and the export chain show no clear break.
04CompressionSlightly deviatingSignalLocal differences can result from editing or export processing.
The download intentionally contains structured sample data only and is not a real forensic report.
Analysis layers
Not one detector. A system of signals.
Each layer answers a different question. Only their combined view reveals which signals align and which conclusions remain open.
Planned
Metadata & file history
EXIF, container and encoding data are checked for breaks, unusual processing steps and inconsistencies.
Limit / Metadata can be altered or removed.
Generative image traces
Local textures, noise patterns, edges and semantic transitions can indicate inpainting or synthetic regions.
Limit / Multiple signals are assessed together.
Face & deepfake analysis
Facial regions are examined for spatial and temporal inconsistencies without claiming identity or intent.
Limit / A technical signal is not proof of wrongdoing.
Video & frame consistency
Frames, motion transitions and temporal artefacts are presented as a timeline rather than reduced to one total score.
Limit / Compression can weaken signals.
Audio & lip synchronization
Speech patterns, audio artefacts and lip movements can contribute an additional evidence layer to the assessment.
Limit / Audio analysis is a separate module.
Localization
Show the location. Not just the verdict.
Suspicious image regions and video segments should be visible directly within the media. Move the slider to compare the sample edit with the source image.
Original:Source image without the inserted person.
Edited:Example of a locally generated image region.
OriginalEdited
Generatively inserted person in the right image region
Illustrative example of localization — not a real case analysis.
Workflow
Upload. Analyse. Understand.
The workflow stays simple. Technical depth appears only when it is relevant to the decision.
Upload
Choose an image or video. Supported formats, file limits and privacy information are shown clearly before upload.
Analyse
Several independent signal layers are examined. Progress and current processing steps remain visible.
Understand
The result separates signals, counter-evidence, provenance and uncertainty in a traceable report.
Use cases
Built for decisions, not effects.
Newsrooms & journalism
Prioritize material before publication and document unresolved questions.
Companies & brands
Review campaign material, testimonials and suspicious publications in a structured workflow.
Platforms & marketplaces
Triage high-risk uploads and support moderation teams with visible evidence.
Agencies & creative teams
Make editing steps traceable and keep Content Credentials visible.
Individuals
Understand suspicious images or videos without specialist technical knowledge.
Method & limitations
Transparency is part of the product.
Generators, editing tools and platform compression evolve constantly. Forenlyx must therefore never imply universal certainty. Models need versioning, regular evaluation and documented known weaknesses.
01Results presented as evidence assessments, not absolute verdicts
02Model version and analysis timestamp included in reports
03Counter-evidence and inconclusive results remain visible
04Manual review for high-impact decisions
Privacy & security
Make the data flow visible.
FORENLYX processes images in memory for each request and does not write the media file to its own persistent storage. A reduced image copy is sent to the configured OpenAI API for visual observations; technical usage data and analysis status are logged in Supabase.
01
File processing
per request, without proprietary persistent media storage
02
External service
reduced visual copy sent to the configured API
03
Usage log
analysis ID, status, file size and quota
04
Public launch
document final retention periods and legal bases
Exportable report
For the next review step.
A professional report documents the file, method, signals, limitations and model version. For teams, it becomes a shared and traceable basis for review.
FORENLYX
Media evidence report
Assessment
Signals of local generative editing.
Manual review recommended01File and hash data
02Analysis layers & model version
03Evidence and counter-evidence
04Limitations & next review steps
Product demo — not a forensic expert report01 / 04
FAQ
Open questions. Clear answers.
01Can Forenlyx prove with certainty that a file was edited using AI?+
No. There is no universally error-free detection method. Forenlyx is intended to combine several technical signals, surface counter-evidence and state uncertainty explicitly. The result is an evidence assessment, not absolute proof.
02What does it mean when no Content Credentials are present?+
Only that no usable provenance record was found. Many cameras, tools and platforms do not yet support Content Credentials or remove metadata during export. Their absence must not be interpreted as proof of manipulation.
03How are uploaded files processed?+
The current application processes images in server memory for the individual request and does not place the media file in its own persistent storage. A reduced image copy is transmitted to the configured OpenAI API for the visual-observation module. Supabase records technical usage data such as analysis ID, status, file size and daily quota. The data terms of the infrastructure and API providers also apply.
04Who is the platform designed for?+
The interface is intended to remain understandable for individuals while supporting professional review workflows. Extended reports, teams, API access and batch processing would be later business features.
05Which analysis is already functional?+
Registered accounts with an active trial or plan can run technical and visual analysis on JPG, PNG and WebP images. The public example analysis remains a clearly labelled simulation. C2PA validation, dedicated deepfake models, and real video, frame and audio analysis are planned as separate modules and are not presented as currently available.
Private product preview
Make digital evidence visible.
Signal / provenance / uncertainty