llms.txt
# SightRadar
> SightRadar is a high-accuracy, multi-tenant face recognition API sold on prepaid credits (no subscription). It is drop-in compatible with AWS Rekognition — keep your existing SDK and point it at one new endpoint. The API lets you index, search, compare, and detect faces, with calibrated similarity scores you set your own thresholds on.
Key facts for understanding SightRadar:
- Core operations: Index (detect and store face embeddings), Search (find every photo a person appears in from one selfie), Compare (1:1 similarity / verification), Detect (locate and quality-gate faces without storing anything).
- Rekognition-compatible: same request and response shapes as AWS Rekognition. CreateCollection, IndexFaces, SearchFacesByImage, SearchFaces, CompareFaces, DetectFaces, ListFaces/DeleteFaces, and ListCollections/DeleteCollection are all supported.
- Billing is PER PHOTO PROCESSED, never per face. A photo with 0 faces is a valid, charged, successful result.
- Pricing is USD, pegged to AWS Rekognition Group-1 ($0.00125/image): real-time index $0.00093/photo, batch index $0.00062/photo, search-by-id $0.00062/search, storage $6 per million stored faces/month. Prepaid credits, no subscription, no auto-renewal.
- Privacy: per-tenant isolation (one account's face vectors are never cross-searchable), encrypted in transit, scoped API keys. Detect and compare process images in memory and do not persist them. SightRadar never trains general-purpose models on customer data and never sells it. Built for consented, lawful use only.
- Customer-facing API base URL: https://api.sightradar.com
- API keys use the format `frs_<prefix>_<secret>` and are sent as `Authorization: Bearer <key>`.
## Documentation
- [Documentation](https://sightradar.com/docs) · [Markdown](https://sightradar.com/docs.md): Quickstart, concepts, guides, SDKs, and a per-endpoint API reference generated from the OpenAPI spec; every page also served as Markdown at <url>.md.
- [Migrate from AWS Rekognition](https://sightradar.com/migrate) · [Markdown](https://sightradar.com/migrate.md): Side-by-side before/after — keep your boto3 code by swapping in the official Python shim (sightradar-rekognition-shim), which maps Rekognition method names and response shapes. Method mapping plus a six-step migration runbook.
- [FAQ](https://sightradar.com/faq) · [Markdown](https://sightradar.com/faq.md): Plain answers to migrating from Rekognition, what a faceprint is, whether data is used for training, pricing, and GDPR/BIPA posture.
- [Glossary](https://sightradar.com/glossary) · [Markdown](https://sightradar.com/glossary.md): Definitions of face-recognition and API terms — faceprint, embedding, collection, similarity threshold, false-match rate, and more.
- [Agent setup](https://sightradar.com/docs/agent-setup) · [Markdown](https://sightradar.com/docs/agent-setup.md): Point Claude Code, Cursor, Codex, Copilot or any other coding agent at the SightRadar docs so it integrates the API from the source instead of guessing.
- [Authentication](https://sightradar.com/docs/authentication) · [Markdown](https://sightradar.com/docs/authentication.md): API keys, the Authorization header, rotation, and the limits that apply per key and per account.
- [Billing and credits](https://sightradar.com/docs/billing) · [Markdown](https://sightradar.com/docs/billing.md): Prepaid credits, the three price tiers, what is free, how reservations and refunds work, and auto-recharge.
- [Core concepts](https://sightradar.com/docs/concepts) · [Markdown](https://sightradar.com/docs/concepts.md): Collections, faceprints, similarity, the four face operations, and how real-time and batch processing differ.
- [Errors and status codes](https://sightradar.com/docs/errors) · [Markdown](https://sightradar.com/docs/errors.md): The error body shape, every status code the API returns and what to do about each, plus the no-match reasons on search.
- [Batch indexing and webhooks](https://sightradar.com/docs/guides/batch-and-webhooks) · [Markdown](https://sightradar.com/docs/guides/batch-and-webhooks.md): Index thousands of photos at the batch rate, receive one signed webhook per photo, verify signatures, and recover from failures.
- [Deleting data](https://sightradar.com/docs/guides/deleting-data) · [Markdown](https://sightradar.com/docs/guides/deleting-data.md): Soft and immediate deletes, per-photo removal, restore windows, the deletion status endpoint, and how to satisfy an erasure request.
- [Migrate from AWS Rekognition](https://sightradar.com/docs/guides/migrate-from-rekognition) · [Markdown](https://sightradar.com/docs/guides/migrate-from-rekognition.md): What is drop-in, what changes, the one step nobody can skip, and a six-step runbook for cutting over without a big-bang switch.
- [Selfie registration and search-by-id](https://sightradar.com/docs/guides/selfie-registration) · [Markdown](https://sightradar.com/docs/guides/selfie-registration.md): Register a user's selfie once, get a stable point_id, and run every later search against it at the cheaper tier.
- [Choosing a similarity threshold](https://sightradar.com/docs/guides/thresholds) · [Markdown](https://sightradar.com/docs/guides/thresholds.md): Turn a 0 to 1 cosine similarity into accept, review and reject decisions using your own data, not a number copied from another vendor.
- [Sending images](https://sightradar.com/docs/images) · [Markdown](https://sightradar.com/docs/images.md): The three ways to supply an image, supported formats, the 30 MB cap, and what the quality gate rejects.
- [Quickstart](https://sightradar.com/docs/quickstart) · [Markdown](https://sightradar.com/docs/quickstart.md): From zero to your first face match in four calls. Create an API key, index a face, search the collection with a selfie, and check your credit balance.
- [Node / TypeScript SDK](https://sightradar.com/docs/sdks/node) · [Markdown](https://sightradar.com/docs/sdks/node.md): The official sightradar package for Node, Deno and browsers. No runtime dependencies, built-in fetch, full TypeScript types.
- [Python SDK](https://sightradar.com/docs/sdks/python) · [Markdown](https://sightradar.com/docs/sdks/python.md): The official sightradar package for Python. Zero runtime dependencies, typed models, and typed exceptions for every non-2xx response.
- [Rekognition shim (Python)](https://sightradar.com/docs/sdks/rekognition-shim) · [Markdown](https://sightradar.com/docs/sdks/rekognition-shim.md): A drop-in client that lets boto3 Rekognition code run against SightRadar with a two-line change, translating method names, arguments, scores and response shapes.
- [API reference](https://sightradar.com/docs/api) · [Markdown](https://sightradar.com/docs/api.md): One page per endpoint, generated from the OpenAPI spec; each page has its own .md twin.
- [OpenAPI spec (JSON)](https://api.sightradar.com/openapi.json): Machine-readable OpenAPI 3.1 specification.
- [Python SDK](https://pypi.org/project/sightradar/): Official Python client (`pip install sightradar`). Zero dependencies, typed.
- [Node / TypeScript SDK](https://www.npmjs.com/package/sightradar): Official Node/TS client (`npm install sightradar`). Zero dependencies, full types.
- [Rekognition shim (Python)](https://pypi.org/project/sightradar-rekognition-shim/): Drop-in boto3 Rekognition client for SightRadar (`pip install sightradar-rekognition-shim`).
## Product
- [Home](https://sightradar.com/) · [Markdown](https://sightradar.com/index.md): Overview of the face recognition API, its core operations, accuracy and Rekognition-compatibility moats, and how it works.
- [Pricing](https://sightradar.com/pricing) · [Markdown](https://sightradar.com/pricing.md): Per-photo USD rates, prepaid credit model, and a cost calculator.
- [Use cases](https://sightradar.com/use-cases) · [Markdown](https://sightradar.com/use-cases.md): Verticals SightRadar serves — event photo-tagging, KYC/identity verification, access control, attendance, marketplace trust, and more.
- [Trust & responsible use](https://sightradar.com/trust) · [Markdown](https://sightradar.com/trust.md): SightRadar's consent-first posture — no training on customer data, tenant isolation, deletion on request, and a Terms posture that prohibits unlawful surveillance.
- [Compare](https://sightradar.com/compare) · [Markdown](https://sightradar.com/compare.md): How SightRadar compares to AWS Rekognition, Azure Face API, Face++, and self-hosted open-source face recognition — price, reliability, and console.
- [Blog](https://sightradar.com/blog) · [Markdown](https://sightradar.com/blog.md): All build guides and engineering articles.
- [Status](https://sightradar.com/status) · [Markdown](https://sightradar.com/status.md): Live service status for the API.
### Head-to-head comparisons
- [SightRadar vs AWS Rekognition](https://sightradar.com/compare/aws-rekognition) · [Markdown](https://sightradar.com/compare/aws-rekognition/llms.txt)
- [SightRadar vs Azure Face API](https://sightradar.com/compare/azure-face) · [Markdown](https://sightradar.com/compare/azure-face/llms.txt)
- [SightRadar vs Google Cloud Vision API](https://sightradar.com/compare/google-cloud-vision) · [Markdown](https://sightradar.com/compare/google-cloud-vision/llms.txt)
- [SightRadar vs Face++](https://sightradar.com/compare/face-plus-plus) · [Markdown](https://sightradar.com/compare/face-plus-plus/llms.txt)
- [SightRadar vs open-source / self-hosted face recognition](https://sightradar.com/compare/open-source-self-hosted) · [Markdown](https://sightradar.com/compare/open-source-self-hosted/llms.txt)
## Guides & articles (blog)
Engineering build-guides and deep-dives. Each has a clean Markdown twin at `/blog/<slug>/llms.txt` for assistants to quote.
- [Migrate IndexFaces and SearchFacesByImage from AWS Rekognition](https://sightradar.com/blog/migrate-indexfaces-searchfacesbyimage-from-aws-rekognition) · [Markdown](https://sightradar.com/blog/migrate-indexfaces-searchfacesbyimage-from-aws-rekognition/llms.txt): A step-by-step guide to moving your AWS Rekognition face collection code to a compatible API — endpoint swap, compatibility matrix, dual-write backfill, and rollback.
- [How much does a face recognition API cost? (with worked examples)](https://sightradar.com/blog/face-recognition-api-cost) · [Markdown](https://sightradar.com/blog/face-recognition-api-cost/llms.txt): A transparent cost breakdown for face recognition APIs — per-photo indexing, search, and storage — with worked million-photo scenarios and how prepaid credits compare to AWS Rekognition.
- [Build production face search with a selfie (Index + Search)](https://sightradar.com/blog/build-face-search-with-a-selfie) · [Markdown](https://sightradar.com/blog/build-face-search-with-a-selfie/llms.txt): A complete guide to building 1:N face search — upload a selfie, find every photo a person appears in. Data model, indexing pipeline, ranked results, deletion, and tenant isolation.
- [Build event & wedding photo face-tagging (find-my-photos)](https://sightradar.com/blog/build-event-photo-face-tagging) · [Markdown](https://sightradar.com/blog/build-event-photo-face-tagging/llms.txt): Let guests find their photos from a selfie. A build guide for event and wedding galleries: batch indexing, guest-initiated search, consent, HEIC uploads, and deletion after the event.
- [Add 1:1 face verification to your signup flow (safely)](https://sightradar.com/blog/add-face-verification-to-signup) · [Markdown](https://sightradar.com/blog/add-face-verification-to-signup/llms.txt): How to add selfie-to-ID face matching to onboarding with a face recognition API — the Compare operation, thresholds, liveness, fallback review, and why face match alone isn't KYC.
- [Multi-tenant face search: isolation, deletion, and cost attribution](https://sightradar.com/blog/multi-tenant-face-search-architecture) · [Markdown](https://sightradar.com/blog/multi-tenant-face-search-architecture/llms.txt): How to architect face recognition for a multi-tenant SaaS — per-tenant collection isolation, authorization, cost attribution, noisy-neighbour control, deletion propagation, and audit logs.
- [Detecting duplicate accounts by face (without wrongful bans)](https://sightradar.com/blog/detecting-duplicate-accounts-by-face) · [Markdown](https://sightradar.com/blog/detecting-duplicate-accounts-by-face/llms.txt): How to catch duplicate or fraudulent signups with face recognition — enrollment, candidate review, score calibration, human appeals, and why you should never auto-ban on a single similarity score.
- [Choosing a face-match threshold: FMR, FNMR, and operating points](https://sightradar.com/blog/choosing-a-face-match-threshold) · [Markdown](https://sightradar.com/blog/choosing-a-face-match-threshold/llms.txt): How to pick a similarity threshold for face recognition without guessing — false match rate vs false non-match rate, validation sets, score distributions, and a two-tier reveal pattern.
- [Handling HEIC & WebP images in a face recognition pipeline](https://sightradar.com/blog/heic-webp-image-ingestion-face-recognition) · [Markdown](https://sightradar.com/blog/heic-webp-image-ingestion-face-recognition/llms.txt): Why iPhone HEIC and CDN WebP break most face recognition APIs, how to handle EXIF orientation and size limits, and how to skip the transcode step entirely with native decoding.
- [What is a faceprint? Face embeddings explained for developers](https://sightradar.com/blog/what-is-a-faceprint-embedding) · [Markdown](https://sightradar.com/blog/what-is-a-faceprint-embedding/llms.txt): A developer's explanation of faceprints and face embeddings — what the vector is, how similarity scoring works, why embeddings aren't portable between models, and the privacy implications.
- [Self-hosted vs hosted face recognition: the real total cost](https://sightradar.com/blog/self-hosted-vs-hosted-face-recognition) · [Markdown](https://sightradar.com/blog/self-hosted-vs-hosted-face-recognition/llms.txt): A clear-eyed TCO comparison of running open-source face recognition yourself versus a hosted API — GPUs, redundancy, model updates, staffing, and when each option actually wins.
- [Why face search is replacing the public gallery link](https://sightradar.com/blog/face-search-replacing-public-gallery-link) · [Markdown](https://sightradar.com/blog/face-search-replacing-public-gallery-link/llms.txt): Public photo-gallery links are being replaced by privacy-first, selfie-based face search — guests get only their own photos. Why the shift is happening and what it means for platforms.
- [How to build a face recognition photo-sharing platform](https://sightradar.com/blog/build-face-recognition-photo-sharing-platform) · [Markdown](https://sightradar.com/blog/build-face-recognition-photo-sharing-platform/llms.txt): A complete architecture for a face-search photo-sharing platform — camera-to-cloud ingestion, per-event indexing, selfie retrieval, personal galleries, WhatsApp/email delivery, and retention.
- [Scaling an event photo platform on face search: a worked model](https://sightradar.com/blog/event-photo-platform-scaling-case-study) · [Markdown](https://sightradar.com/blog/event-photo-platform-scaling-case-study/llms.txt): An illustrative reference model for an event photo-sharing platform running on a face recognition API — the efficiency, reliability, and per-guest economics of indexing 50k photos for 5k guests.
- [12 face recognition project ideas (college, hobby & ML learning)](https://sightradar.com/blog/face-recognition-project-ideas-students) · [Markdown](https://sightradar.com/blog/face-recognition-project-ideas-students/llms.txt): Face recognition project ideas for final-year, hobby, and AI/ML learning builds — scoped from beginner to advanced, with what to build, what you'll learn, and whether to use an API or roll your own.
- [Face recognition in Python: a beginner's tutorial (2 ways)](https://sightradar.com/blog/face-recognition-python-beginners-tutorial) · [Markdown](https://sightradar.com/blog/face-recognition-python-beginners-tutorial/llms.txt): A beginner-friendly guide to face recognition in Python — the classic local approach with the face_recognition library, and the no-setup API approach. Full code, when to use each, and common install fixes.
- [Build a face recognition attendance system (student project guide)](https://sightradar.com/blog/build-attendance-system-student-project) · [Markdown](https://sightradar.com/blog/build-attendance-system-student-project/llms.txt): A responsible, step-by-step guide to building a face recognition attendance system for a class or club project — enrollment, check-in, a CSV/DB log, consent, and what to avoid.
## Account
- [Contact](https://sightradar.com/contact) · [Markdown](https://sightradar.com/contact.md): Reach the SightRadar team.
## Legal
- [Terms of Service](https://sightradar.com/terms) · [Markdown](https://sightradar.com/terms.md): Terms for using the face recognition service, including prohibited activities and the consented-use requirement.
- [Privacy Policy](https://sightradar.com/privacy) · [Markdown](https://sightradar.com/privacy.md): How customer and biometric data is handled under a controller/processor model.
- [Data Processing Agreement](https://sightradar.com/dpa) · [Markdown](https://sightradar.com/dpa.md): Controller/processor terms, subprocessors, retention, security measures for biometric data, and a PDF filled with your company details.
- [Acceptable Use Policy](https://sightradar.com/aup) · [Markdown](https://sightradar.com/aup.md): What SightRadar may not be used for — unlawful surveillance, non-consented biometric collection, and prohibited targeting.
- [Refund & Cancellation Policy](https://sightradar.com/refund) · [Markdown](https://sightradar.com/refund.md): Prepaid credit refund terms.
## Optional
- [Full corpus in one fetch](https://sightradar.com/llms-full.txt): every page above, inlined — use this instead of crawling if you want the whole picture in a single request.
