Signature API
The Signature API compares a questioned signature with one to five reference signatures from the same person and returns the result as an explainable confidence score from 0 to 100. The decision is never left to a single model: 23 independent models vote, stroke-curve biometrics show which feature each vote rests on, and if you ask for it the result becomes a PDF report fit to be put before a court.
How It Works
A comparison request is sent to the POST /v2.4/signature/comparisons endpoint. The images are preprocessed (background cleanup, skew correction, resolution normalisation), then the stroke skeleton is extracted. Each model reaches its own decision from that skeleton and the raw image; the consensus layer weights the votes and produces a single score. The result is ready within a few seconds and arrives through the signature.compared webhook event or by polling.
Every comparison costs $0.40 in credits. The charge comes out of your team's shared credit balance; the PDF report costs nothing extra.
Pages in This Section
| Page | What it covers |
|---|---|
| comparison-requests | The request body, how many references to send, ways to get the result and the response shape. |
| confidence-score | How the votes of 23 models become one score, the score bands and explainability. |
| court-report | The sections of the court report, how to request it and how long it is kept. |
| biometric-analysis | Stroke curve, pressure estimate, speed, tremor, slant and proportion features. |
For a first integration, start with Signature Comparison Requests; read Confidence Score Algorithm before you show the score to anyone.
The Signature API is built to speed up and justify forensic document examination. For decisions with legal consequences, use the score together with the assessment of a qualified handwriting or forensic document examiner.