TurboLens

TurboLens is a content detector tool. A REST JSON API that scores documents and images for local pixel manipulation (splicing, copy-move, retouching) and returns a verdict, tampered-pixel share, region coordinates, and heatmap. An early-access layer screens for AI-generated and AI-edited media.

TurboLens is a visual forensics API for teams that need to know whether an image has been manipulated before it influences a decision. The service analyzes pixel structure to find local tampering: splicing, copy-move edits, retouching, and altered fields on documents such as payslips, bank statements, passports, and identity cards. Each submission returns a real or forged verdict, the proportion of tampered pixels, coordinates for suspicious regions, and a heatmap that shows a human reviewer where an edit may sit. A second, separate layer screens for synthetic media, estimating whether generative AI created or altered the image and returning a three-state result with a probability score. The core design choice is that forgery detection and AI-image detection stay as two distinct signals rather than one blended score. Forgery detection is production ready; the synthetic-media layer is early access. Practically, this means an operations team can place one REST call immediately after upload and before OCR, extraction, or approval. Clean files continue down the existing pipeline automatically, while flagged images are rejected, escalated, or queued for manual review according to thresholds the organization sets itself. TurboLens also handles document intelligence after the authenticity check: classifying documents and pulling structured fields from invoices, receipts, passports, national IDs, contracts, tax forms, and customs paperwork, with an emphasis on Southeast Asian layouts and multilingual scripts. Vietnamese invoice, contract, and customs extraction are called out as named capabilities. The tool fits KYC and onboarding, lending underwriting, and insurance claims, where manipulated evidence is a known risk and review capacity is limited. It is a developer-facing product sold as an API, not a desktop app, so it suits engineering and risk teams building their own review tooling rather than analysts wanting a point-and-click interface. Compared with general OCR vendors, TurboLens leads with authenticity evidence; compared with pure deepfake detectors, it keeps the extracted data workflow in the same stack.