face recognition API

Rekognition-compatible face recognition API.

Index, search, compare, and detect faces at scale, billed per photo processed — with confidence scores you can threshold on.

Already using AWS Rekognition? Keep your boto3 code — change two lines and ship.

Start free — no card Try the live demo

Pay only for what you use · billed per photo, never per face · no minimums.

Why teams switch

PRICED PER PHOTO

$0.00062 a photo, batch

38% under AWS Rekognition's first US tier ($0.0010). real-time $0.00093. run the numbers yourself

GDPR READY

A real DPA, and deletion that actually works

Article 28 processor terms, Article 15/17 subject requests, never trained on your data

Run your own selfie against the live demo below.

don't take our word for it

See it recognize faces — with the JSON.

Search a gallery from one selfie, verify two faces 1:1, or detect and quality-gate. Every tab shows the actual Rekognition-shaped response, so you can see the engine is real and the API is compatible.

Query selfie

4 matches · 41ms

Collection (8 photos)

0.95

0.41

0.96

0.38

0.44

0.92

0.36

0.97

API responseRekognition-compatible

// SearchFacesByImage response
{
  "SearchedFaceConfidence": 99.81,
  "FaceMatches": [
    { "Similarity": 98.1, "Face": { "FaceId": "b3f1…a7" } },
    { "Similarity": 97.4, "Face": { "FaceId": "c2a4…b9" } },
    { "Similarity": 96.7, "Face": { "FaceId": "7e51…62" } },
    { "Similarity": 97.2, "Face": { "FaceId": "9a04…e2" } }
  ]
}

Faces shown are AI-generated — these people don't exist. The matched shots are the same generated identity with real appearance changes (glasses, clean-shaven, aging, different lighting and settings). Scores are representative. Measured accuracy, with methods and limits, is on the benchmarks page.

why teams pick SightRadar

Two reasons it sticks.

Scores you can threshold

Calibrated similarity scores you set your own threshold on, and every search match labelled auto (safe to reveal) or review (confirm first). Measured results, methods and limits are on the benchmarks page.

Rekognition-compatible API

Every Rekognition face operation has a SightRadar equivalent. On Python, the official shim keeps your boto3 IndexFaces / SearchFacesByImage / CompareFaces calls and Rekognition-shaped responses with a two-line change. Other languages call the REST API.

migrate without a rewrite

Keep your code. Change two lines.

SightRadar mirrors AWS Rekognition's face operations. On Python, the official sightradar-rekognition-shim keeps Rekognition's method names, arguments and response shapes: swap the import and the constructor, and your existing calls keep working. From any other language, call the REST API.

Migration guide →SightRadar vs AWS Rekognition →

Supported operations

CreateCollection
IndexFaces
SearchFacesByImage
SearchFaces
CompareFaces
DetectFaces
ListFaces / DeleteFaces
ListCollections / DeleteCollection

# pip install sightradar-rekognition-shim

# was: import boto3
from sightradar_rekognition_shim import client

# was: rek = boto3.client("rekognition")
rek = client(api_key="frs_...")

# your existing calls, unchanged:
rek.search_faces_by_image(
    CollectionId="guests",
    Image={"Bytes": selfie},
    FaceMatchThreshold=80,
)

four moves, one clean API

Everything you need to see faces.

Index

Detect every face in a photo and store its embedding in your collection.

Search

Find every photo a person appears in from a single selfie.

Compare

Similarity between two faces — verify identity 1:1.

Detect

Locate and quality-gate faces without storing anything.

honestly, it's three steps

Set it once. Then just call it.

Grab a key

Open the console, add credits, copy your API key.

Index your faces

POST photos to a collection. We detect, embed, and store every face.

Search & verify

Send a selfie or compare two faces. Same calls you already know.

handle faces responsibly

Sensitive data, treated that way.

Face data is biometric and sensitive. We keep our practices deliberately narrow — and write them down in plain language.

You control retention

Detect and compare process images in memory and don't persist them. Indexed faces live in collections you can delete any time, via API or console.

Isolated & encrypted

Each account's face vectors are namespaced and never cross-searchable. Encrypted in transit, scoped API keys, per-tenant isolation.

Never trained on your data

We process your data only to return your results. We do not pool it or use it to train general-purpose models, and we never sell it.

Built for consented use

You're the data controller. Our Terms require a lawful basis and consent — and prohibit unlawful surveillance or covert tracking.

Read the full Privacy Policy and Terms.

and yes — it costs less

Billed per photo, not per face.

Pay only for what you use, billed per photo processed (never per face) — no surprise bills. Control-plane calls are free.

Real-time — $0.00093 per photo processed
Batch — $0.00062 per photo processed
Search by face-id — $0.00062 · storage $6 / M / mo

See full pricing & calculator

Batch indexing

$0.00062

per photo processed

Pay only for what you use — no subscription, no minimums, no surprise bills. See the calculator for a side-by-side with AWS Rekognition.

Start with pay-as-you-go

Ship face search this week.

Free to start, pay only for what you use. Your first match in minutes — no rewrite.

Get your API key — free