Amazon Rekognition

What it does

Amazon Rekognition takes in images and video, either stored in Amazon S3 or streamed from Kinesis Video Streams, and returns analysis results such as detected objects, scenes, faces and facial attributes, text, celebrities, unsafe content, and video segments. Output is delivered through API responses that your application consumes.

People typically add it to their own applications through its APIs rather than using a standalone app. Common uses include content moderation, identity verification with facial comparison and liveness checks, media analysis to find key video segments, and smart alerts for connected cameras. Custom Labels lets you train a model to recognize your own objects, such as brand logos, from as few as 10 images.

Rekognition suits developers and businesses building computer vision features who do not want to build or maintain their own ML models. It requires an AWS account, and billing is pay-as-you-go per image, per minute of video, or per storage unit, with no up-front commitment; the Free Tier covers limited monthly usage for 12 months. Custom model training and inference bill by hour, and costs can grow quickly at high volume.

What it can do

• Label detection for objects, scenes, activities, and landmarks
• Face detection, analysis, comparison, and search
• Text detection in images and video
• Content moderation for unsafe or unwanted content
• Celebrity recognition
• Video segment detection such as shots and credits
• Custom Labels training for organization-specific objects

Works with

• Amazon S3 (stored video and image analysis)
• Amazon Kinesis Video Streams (streaming video events)

Example workflows

Moderate uploaded images

Input: user-uploaded images sent to the DetectModerationLabels API. Steps: call the API per image, read returned moderation labels, and route flagged items to review. Output: a list of labels with confidence scores for each image.

Find key segments in a stored video

Input: a video file stored in Amazon S3. Steps: start a segment detection job on the video, poll for results, and read the returned segment types. Output: timestamps for shots, black frames, and credits that editors can use for ad insertion or content operations.

Before you choose

• Cost depends on usage tier and API group; running multiple APIs on one image bills each call separately, and face metadata storage is billed monthly.
• Custom Labels training and inference are billed per hour, and training jobs over 72 hours are terminated automatically.

Listed under

• Coding
• API
• Computer vision
• Computer-vision
• Content moderation
• Content-moderation
• Developers

Get started

Sign-up is through the official AWS getting started flow.

1. Create or sign in to an AWS account and open the Rekognition console
2. Choose the APIs you need (for example DetectLabels or DetectModerationLabels) and review the pricing groups for them
3. Store input images or videos in Amazon S3, or set up a Kinesis Video Stream for streaming analysis

Image analysis is tiered by monthly volume (Group 1 and Group 2 APIs priced separately, Image Properties separate). Video is billed per minute of processed video. The Free Tier lasts 12 months and includes 1,000 images per month per group, 60 minutes of video analysis per month, and 2 training and 1 inference hour per month for Custom Labels; a 6-month free plan and up to $200 credits apply to new accounts from July 15, 2025. Pricing examples on the page are illustrative and not general rates.

Source: aws.amazon.com/rekognition/pricing, checked 28 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Image recognition software, ML image analysis, and video analysis – Amazon Rekognition – AWS ↗aws.amazon.com
• Image recognition software, ML image analysis, and video analysis – Amazon Rekognition pricing – AWS ↗aws.amazon.com

Company website

aws.amazon.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.