Overview

SightRadar is a production-grade face recognition API built for developers who need to index, search, compare and detect faces at scale. Instead of shipping a generic vision model and asking you to build the matching layer yourself, it exposes the four operations identity products actually need — index, search, compare, detect — and returns confidence scores you can threshold on for your own risk tolerance.

Its headline proposition is migration cost. The API is drop-in compatible with AWS Rekognition: the request and response shapes match, so teams already on Rekognition can keep their existing SDK, keep their code, and simply point it at one new endpoint. For teams starting fresh, the API works the same way, with docs, a live demo and a console available. Pricing is usage-based and starts from $0.00062 per photo, roughly half the AWS Rekognition rate, and you can start free without a credit card.

Key Features

• Face indexing — enroll faces into collections so they can be searched later, matching against your own data rather than a public corpus.
• 1:N face search — find a person across thousands of photos starting from a single selfie, the classic "who is this and where else do they appear" query.
• 1:1 face compare / verification — compare two faces and get back scores, with the decision threshold set by you rather than hard-coded by the vendor.
• Face detection — locate faces in an image and return their positions, useful for quality gating, cropping and pre-processing before indexing.
• Confidence scores everywhere — every result carries a score, so you can tune between false accept and false reject rates per product (e.g. strict for payments, looser for photo tagging).
• AWS Rekognition compatibility — identical request and response shapes, so an existing Rekognition integration becomes an endpoint swap rather than a rewrite.
• Graceful zero-face handling — a photo with 0 faces returns a valid empty result instead of an error, which keeps batch pipelines from failing on edge cases.
• Per-photo pricing — billed per photo, never per face, from $0.00062 per photo, with no minimums.
• Fast onboarding — free start with no card, a live browser demo for validating accuracy on your own images, and full docs.

Target Users

• Backend and platform engineers building identity, verification or media-search features who want an API rather than a model to host.
• Teams already on AWS Rekognition who want lower cost or a second source without rewriting their integration.
• Consumer and marketplace product builders — dating and social apps, event photography, photo libraries, marketplace trust-and-safety screening, Know Your Customer flows.
• Startups and MVP teams that need high accuracy without infrastructure or volume commitments, thanks to per-photo billing and no minimums.
• Security and access-control developers implementing 1:1 verification or walk-in/selfie check-in flows.

Unique Selling Points

1. Endpoint-swap migration. Same request and response shapes as AWS Rekognition — the switching cost is measured in minutes, not sprints.
2. Cheaper, simpler economics. From $0.00062 per photo, about half the AWS Rekognition rate, and billed per photo rather than per detected face, so group photos and crowded scenes don't multiply your bill.
3. Thresholds belong to you. You receive scores and choose your own cut-offs, so the same API can power a lenient photo-tagging feature and a strict verification flow.
4. Zero-face photos are normal. Empty results are valid, which removes a common source of failed jobs in production pipelines.
5. Frictionless evaluation. Start free, no card, no minimums, and test accuracy in a live demo before committing to anything.

Pros and Cons

Pros

• Near-zero migration cost for existing AWS Rekognition users thanks to matched request/response shapes.
• Transparent, usage-based pricing at roughly half the incumbent rate, with no minimums and no card required to start.
• Full operation set in one API: index, search, compare and detect, so you don't stitch multiple vendors together.
• Confidence scores on results plus user-defined thresholds give real control over precision/recall trade-offs.
• Sensible edge-case behaviour — a photo with no faces is a result, not an exception.
• Live demo and documentation make it easy to validate on your own images before integrating.

Cons

• Accuracy and cost comparisons on the page are vendor-stated; no independent third-party benchmark is published there.
• The page doesn't describe self-hosted, on-premises or region-specific data-residency deployment options, which are often hard requirements in biometric projects.
• Compliance obligations for biometric data (consent capture, retention limits, GDPR/BIPA-style regimes) remain entirely with the customer — the API provides capability, not compliance.
• Enterprise specifics such as SLAs, support tiers and formal certifications are not covered on the landing page; confirm them on the pricing and docs pages before committing to a regulated workload.
• Overlap is likely on the roadmap for teams that also need broad general-purpose computer vision (labels, moderation, text-in-image) beyond faces.