— Products
One engine. A growing set of instruments.
One product is live today, as a demo build labeled as one. Three more are in development on the same calibrated engine, and they’re labeled too. Nothing on this page carries a number our benchmark hasn’t produced.
— Available today
What you can run right now
For labs, OSINT & private users
AI image detection & forensic reports
A detector fine-tuned for current generators scores your image and returns a verdict at a capped false-positive rate: 5% for triage, 1% for evidence. Thresholds are calibrated per condition, detection power is quoted under the image’s estimated condition, and every limit is stated on the report. Chain-of-custody hashes cover the exhibit and the report itself.
— In development
Where the engine goes next
Three vertical products being built on the same calibrated engine. No launch dates and no invented accuracy claims. When they ship, their reports will state their limits the way today’s do.
For law firms, PIs & HR investigations
Evidence review for legal teams
Court-ready review of disputed images: chain-of-custody hashing on receipt, analysis under a named calibration run, and error rates stated under realistic degradations. That is the kind of claim that survives cross-examination where a vendor’s “99% accurate” dies. Per-case pricing, built on the same engine as today’s reports.
For insurers & TPAs
Claim-photo verification
Screening for claim photos: AI-generation likelihood, edit and splice signals, EXIF consistency, and duplicate detection against prior claims; recycled photos are fraud’s cheapest trick. Built for images that arrive through app and email re-encode chains, because that is exactly what our benchmark measures.
For journals & research-integrity offices
Scientific-image integrity
Submission screening for manipulated, duplicated, or AI-generated figures. Batch scoring across a manuscript’s images feeds a ranked review queue. The pipeline flags and explains; your editors decide.
One of these your problem today?
We’re looking for design partners, not signups. Tell us about your workload and the re-encode chain your images survive before they reach you.
products@vestigeforensics.com
— On the roadmap
Further out, labeled honestly
Two new modalities, programmatic access, and a standing commitment to model freshness. Roadmap means exactly that: no launch dates, and no accuracy claims until a benchmark backs them.
New modality
AI video detection
Extending the calibrated methodology to generated and manipulated video: capped false-positive rates, condition-aware thresholds, limits printed on the report. The benchmark comes first; there are no numbers to quote until it exists.
New modality
AI voice detection
The same benchmark-first discipline applied to synthetic speech, calibrated under the degradations voice actually travels through: codecs, calls, re-uploads. No dates, and no detection rates until a benchmark produces them.
The live image engine
API & batch access
Programmatic access to the engine behind today’s reports is live: bearer-key auth, single-image scoring, and the same versioned report JSON. Keys are included with your plan and spend the quota you already have. Batch scoring in one request is still in engineering.
Read the API docs →Standing commitment
Continuous model refresh
New generators erode detectors within months; our benchmark exists to measure exactly that decay. So the primary detector is retrained and recalibrated against current generators, and every report names the calibration run it was scored under.
Want a ping when any of these move? hello@vestigeforensics.com
Meanwhile, the detector is live
Run an image through the demo and read the full report: thresholds, limits, hashes and all. The products above will be announced the same way, with their numbers.