Detect pixel-level edits and screen AI-generated or AI-edited images before they enter KYC, lending, insurance, and document-processing workflows. Continue with specialized OCR and structured extraction when the same documents need to move downstream.
Verdict
Manipulation detected
Tampered
Flagged
Regions
01
"result": "forged",
"review": true,
"heatmap": "ready"
Two detection layers
Image forgery detection examines an existing image for local manipulation. AI image detection screens whether generative AI created or edited the image. TurboLens keeps the results separate so teams can route each risk through the appropriate review workflow.
| Comparison | Image Forgery Detection | AI-Generated & AI-Edited Detection |
|---|---|---|
| What it detects | Local manipulation in an existing image, including splicing, copy-move edits, retouching, and altered fields. | Signs that an image was created by generative AI or changed with generative editing tools. |
| Decision output | Real or forged verdict, tampered-pixel proportion, suspicious-region coordinates, and a visual heatmap. | Three-state screening result and an AI probability score for review workflows. |
| Workflow role | Locate suspected edits before OCR, extraction, approval, or manual review. | Screen for synthetic-media risk and route uncertain results for review. |
| Availability | Production ready | Early access |
Find evidence of local manipulation in an existing image. TurboLens analyzes pixel structure for splicing, copy-move edits, retouching, and altered fields.
Decision output
Real or forged verdict, tampered-pixel proportion, suspicious-region coordinates, and a visual heatmap.
Screen images for synthetic origin or generative edits, including fully AI-generated documents and images modified with AI editing tools.
Decision output
Three-state screening result—AI-generated, needs review, or likely human-made—plus an AI probability score.
One call before extraction
TurboLens adds a visual-forensics decision before the rest of your workflow. Your team controls the threshold and whether a flagged image is rejected, escalated, or queued for human review.
{
"verdict": "forged",
"tampered_proportion": 0.124,
"review_recommended": true,
"suspicious_regions": 1
}Send a document page or image to the API before OCR and downstream processing.
Run pixel-level forgery analysis and, when enabled, synthetic-media screening.
Continue clean files automatically and send flagged results to your review workflow.
More than visual forensics
After screening image authenticity, use TurboLens to classify documents, extract structured fields, and process complex Southeast Asian layouts and scripts through API-first workflows.
Explore Document IntelligenceExtract supplier, tax, line-item, and payment fields from regional invoice and receipt formats.
View capabilityCapture structured identity fields from Southeast Asian national IDs, passports, and multilingual scans.
View capabilityTurn contracts and legal documents into structured clauses, parties, dates, and reviewable fields.
View capabilityProcess tax forms, customs documents, and mixed-language layouts designed for Southeast Asian workflows.
View capabilityBuilt for high-trust decisions
Visual forensic signals help operations teams focus manual review where it matters while genuine submissions continue through the existing workflow.
KYC & onboarding
Screen passports, national IDs, driving licences, and supporting documents for photo swaps, edited fields, and synthetic creation.
Lending
Find suspicious edits in payslips, bank statements, tax forms, and other income evidence before a credit decision is made.
Insurance
Check receipts, invoices, medical documents, and claim photos for local tampering or generative edits before payout review.
Keep your OCR, face matching, watchlist screening, and case-management tools. TurboLens adds an authenticity signal through a REST endpoint before extraction.
Start with your own samples
See how altered and clean images are scored in your workflow.
Review the integration processCommon questions
Image forgery detection analyzes an existing image for evidence of local manipulation, such as splicing, copy-move edits, retouching, or altered fields. TurboLens returns a real-or-forged verdict, the proportion of suspicious pixels, region coordinates, and a heatmap that helps reviewers see where an edit may have occurred.
Forgery detection looks for edits made to an existing image, while AI-generated image detection estimates whether generative AI created or modified the image. These are separate forensic questions, so TurboLens treats them as two complementary screening layers rather than one interchangeable score.
The synthetic-media service is being developed to screen both fully AI-generated images and images changed with generative editing tools. It is currently offered as early access, with a three-state result and probability score designed to support—not replace—human review and business rules.
Most teams place visual forensics immediately after upload and before OCR, extraction, or approval. Clean submissions continue through the existing pipeline. Flagged images can be rejected, escalated, or sent to a manual-review queue according to the organization’s own thresholds and risk policy.