●Detection · Edits & tampering
Edited photo detection.
Find out if a real photo was changed: objects added or removed, damage exaggerated, numbers altered or the background swapped.
●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.
01
Detect
Every file is scanned by independent models for images, video and faces, in the web tool, the extension or by API.
4 / 4 signals corroborateCalibrated likelihood
02
Explain
See which signals fired and how strongly. No single model decides; every result comes with its reasons.
Result: AI-generated · 0.92<3s
03
Decide
A plain-language result, Real or AI-generated, with a likelihood and the reasons behind it.
●Coverage
What counts as an edit.
Real photos, changed after capture.
Changes
Common targets
●Method
Where an edit gives itself away.
Each check runs as an independent model. Results corroborate before a result is returned.
01
Splice boundaries
Edges where pasted content meets the original photo.
02
Cloned regions
Duplicated patches used to hide or add objects.
03
Noise patterns
Sensor noise that changes across the frame.
04
Edited numbers
Digits and text that were re-drawn on odometers, tags and labels.
05
Background swaps
A real person placed on a generated or different background.
06
Metadata
Date and software inconsistencies in EXIF.
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.
LiveVideo
Face swaps across 32 frames and fully generated video from Runway, Kling, Veo and Pika.
LiveVoice
Cloned and synthesised speech in calls, notes and recordings.
LiveWhat you get
Each file is marked Real or AI-generated, with the likelihood, signals and audit ID behind it.
One answer●When it matters
Wherever a photo is evidence.
Damage added, exaggerated or removed.
Odometers and serial numbers altered.
Price tags and product photos changed.
Edited certificates and documents.
Photos edited before they reach an exhibit list.
Objects removed from news photos.
●How it works
Three steps. One result.
Upload a photo
JPG, PNG or WebP up to 50 MB.
Edit analysis
Splicing, cloning, noise and metadata checks run in parallel.
Result with regions
See which parts were changed and why.
●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
AvailableUpload in the web app. Images, video and voice, up to 50 MB. Full results in seconds.
Chrome extension
AvailableScan any image or video from inside the tools your team already uses. No download, no upload.
API
SoonSingle scans, batch jobs and webhooks. Structured JSON with result, likelihood, signals and audit ID.
curl -X POST https://witness.vision/api/v1/detect \ -H "Authorization: Bearer $WITNESS_KEY" \ -H "Content-Type: application/json" \ -d '{"media_url": "https://…/odometer_reading.jpg"}' → 200 OK · 412ms result: ai_generated · 0.91
●Why Witness
Built to be checked.
We say what the system does, what it returns and where its limits are.
Detection built per signal
Specialized models per signal, so you get more coverage and fewer false positives than one classifier.
Reasons, not just a score
Every result shows the signals behind it and where in the file they fired.
Analysed, then gone
Media is never stored, shared or used for training.
Provenance, when it exists
Reads Content Credentials from capture device to current file.
●Explore
One engine. Every surface.
The same detection, result and audit trail across every team that handles media.
Enterprise
Insurance
Screen damage photos and injury documentation for AI generation before they reach your adjusters.
Explore →Enterprise
Legal
Know which exhibits warrant forensic review.
Explore →Enterprise
Fraud detection
Detect AI-generated images, face-swapped video and cloned voices inside your fraud stack.
Explore →Enterprise
E-commerce
Screen damage, missing-item and not-as-described evidence for AI generation and reuse, before the refund is issued..
Explore →Product
Chrome extension
Scan media on any site, social feeds, news, claims portals, case files, and get a result in the side panel..
Explore →Is an edited photo the same as an AI-generated one?
No. An edited photo started as a real capture. Witness reports edits and generation separately.
Will a beauty filter be flagged?
Light filters are not treated as manipulation. Heavy edits and AI enhancement are flagged with a lower severity.
Can it read edited numbers?
It shows where digits were altered. It does not guess the original value.