AI / ML
Amazon Rekognition
Amazon Rekognition analyzes images and stored or streaming video for labels, moderation, text, faces, celebrities, people paths, and custom labels through managed computer-vision APIs.
Explore pricing models, common use cases, infrastructure support, and the AWS services that commonly work with Amazon Rekognition.
Amazon Rekognition pricing and cost programs
Pricing model: Image and video analysis usage
- On-Demand
- Available
- Reserved Instances or reserved capacity
- Not applicable
- Savings Plans
- Not applicable
- Spot
- Not applicable
Billing dimensions: Images processed · Video minutes · Face-vector storage · Custom Adapter training
Programs and modes: Image APIs · Stored faces · Video analysis · Custom Labels
API family, stored metadata, and media duration determine charges.
Free Tier: Available — verify current offers
Pricing reviewed 2026-07-25. Reviewed against the linked official AWS pricing page. Recheck regional rates and program terms before purchase.
Official AWS sources reviewed 2026-07-21.
Why implement Amazon Rekognition?
- Provides managed image and video analysis without training or operating general-purpose vision servers.
- Supports synchronous image calls and asynchronous stored-video analysis, with streaming-video patterns for supported tasks.
- Returns confidence, geometry, timestamps, and identifiers that applications can combine with task-specific rules and review tools.
How to implement Amazon Rekognition
- Define the legitimate purpose, consent and retention rules, prohibited uses, population and environment, consequence of errors, and representative evaluation dataset before selecting an API.
- Use encrypted S3 or supported streams, least-privilege roles and collections, asynchronous notifications for video, idempotent result handling, and lifecycle deletion for source and derived biometric data.
- Calibrate thresholds per task and risk, present source evidence and confidence to trained reviewers, record overrides, and periodically reevaluate accuracy across lighting, devices, demographics, and changing content.
Amazon Rekognition best practices
- For sensitive face workflows, use high similarity thresholds and human review; AWS suggests 99 percent confidence for use cases where false matches have serious consequences.
- Do not use inferred gender expression or emotion to determine access, rights, benefits, employment, or other consequential outcomes, and do not treat face comparison as identity proof by itself.
- Minimize biometric data, encrypt and tightly restrict collections, document consent and deletion, monitor error and override rates, and pause automation when drift or disparate performance appears.
Amazon Rekognition use cases and server impact
- Media tagging and moderation assistance
- Image text and object detection
- Human-reviewed face search and comparison where lawful
Replaces computer-vision inference fleets, while lawful-purpose review, biometric governance, thresholds, human adjudication, data lifecycle, and downstream decisions remain customer responsibilities.
Official implementation resources
Commonly paired AWS services
- Amazon Simple Storage Service — Object storage
- AWS Lambda — Run code without servers
- Amazon Kinesis — Real-time streaming
- AWS Key Management Service — Key management
- Amazon CloudWatch — Metrics & logs
- Amazon API Gateway — Managed APIs
Architecture patterns using this service
- Process images and media after an Amazon S3 upload — Take a presigned Amazon S3 upload through EventBridge and SQS to Lambda, Fargate, or AWS Batch workers, write derived assets to a separate bucket, and serve them through CloudFront.