Amazon Rekognition Pricing Calculator - Image, Video, Custom Labels, Face Liveness & Custom Moderation
Estimate Amazon Rekognition costs for Group 1 and Group 2 image analyses, Image Properties, face and user vector storage, stored and streaming video, Custom Labels resource-hours, Face Liveness checks, and Custom Moderation adapters.
Frequently Asked Questions
Does Amazon Rekognition charge per image or per API call?
Per API call. AWS states that running multiple APIs against a single image counts as processing multiple images. Sending one photo through DetectLabels, DetectText, and IndexFaces is three billed analyses, not one. This calculator asks for API image analyses for that reason.
What is the difference between Group 1 and Group 2 Rekognition APIs?
Group 1 covers the face and user collection operations: AssociateFaces, CompareFaces, DisassociateFaces, IndexFaces, SearchFaces, SearchFacesByImage, SearchUsers, and SearchUsersByImage. Group 2 covers detection: DetectFaces, DetectLabels, DetectModerationLabels, DetectProtectiveEquipment, DetectText, and RecognizeCelebrities. Each group has its own volume tiers, and the two never share a tier counter. Their top tiers differ too: Group 1 bottoms out higher than Group 2.
Is Image Properties included in Group 2 pricing?
No. Image Properties is called through DetectLabels with the IMAGE_PROPERTIES parameter and is priced on its own schedule. Calling DetectLabels with both GENERAL_LABEL and IMAGE_PROPERTIES incurs both charges.
How does Amazon Rekognition meter stored and streaming video?
Stored video analysis, shot detection, technical cue detection, streaming video events, and live stream processors use separate per-minute meters. Running multiple analyses against the same footage bills each analysis separately. For streaming events, AWS analyzes up to 120 seconds per event; the calculator can derive processed minutes from the event count and seconds per event.
Why can Custom Labels bill more hours than the training actually took?
Custom Labels bills resource-hours, not elapsed time. AWS may run several compute resources in parallel, so a job that finishes in 4.5 hours on two resources bills 9 hours. Provisioned inference works the same way and keeps billing while the model is running even if it processes no images, so deprovision between batches.
Does a Custom Moderation adapter also incur the standard moderation charge?
The official Amazon Rekognition pricing examples for Custom Moderation bill adapter analyses only at the Custom Moderation rate, with no separate Group 2 DetectModerationLabels line. This calculator follows those examples and charges the adapter rate once per analysis.
Can new AWS accounts still use Streaming Video and Bulk Image Analysis?
No. AWS documents that Streaming Video and Bulk Image Analysis are no longer available to new customers, effective 30 April 2026. Accounts that used those features in the previous 12 months keep access, so the calculator still prices them and flags them clearly.
Related tools
CloudBurn
Catch AWS cost mistakes before they ship.
scan against Terraform and CloudFormation before deploy, then run discover against live AWS to find the waste that is already burning.Read the docs to learn how it works.