Enterprise face recognition
Amazon Rekognition
Amazon Rekognition is the image and video analysis service of Amazon Web Services. Its face APIs detect faces, compare two faces, and search a customer's own face collection by image. It is a building block sold to developers by the image, not a web face search. Since June 2020 Amazon has barred police use of its face recognition, a moratorium it extended indefinitely in May 2021.
Updated November 6, 2026 · 3 min read · reviewed November 6, 2026
- $1.00
- Per 1,000 face images, first tier
- 1,000
- Free images a month per group for 12 months (older accounts)
- 2020
- Police moratorium began; extended indefinitely in 2021
AWS Rekognition pricing, Nov 6, 2026
AWS free tier
Reuters, May 2021
What are the face APIs?
Rekognition is billed per image analysed, and running several APIs on one image counts as several images (AWS pricing, checked November 6, 2026). The face APIs are DetectFaces (find faces and landmarks), CompareFaces (one face against another), IndexFaces (add faces to a collection), SearchFacesByImage and SearchFaces (find the closest faces in a collection) and the user-vector APIs that group several faces of one person. A collection is a customer's own repository of face metadata; Amazon runs no collection of the web.
- Step 1
IndexFaces
A customer submits images; Rekognition stores face vectors in the customer's collection.
- Step 2
SearchFacesByImage
A new image is compared with the vectors in that collection.
- Step 3
Similarity threshold
Matches above the customer's chosen threshold come back with a similarity value.
- Step 4
Customer decides
What happens next is the customer's application logic, not Amazon's.
| Volume per month | Group 1 (CompareFaces, IndexFaces, SearchFacesByImage) | Group 2 (DetectFaces) |
|---|---|---|
| First 1 million images | USD 0.0010 per image | USD 0.0010 per image |
| Next 3 to 4 million | USD 0.0008 | USD 0.0008 |
| Next 30 million | USD 0.0006 | USD 0.0006 |
| Beyond | USD 0.0004 | USD 0.00025 |
| Face vector storage | USD 0.00001 per vector per month | n/a |
Why is it on a wiki about face search?
Because it is what a face search engine is made of. A consumer facial recognition search combines a detector, an embedding model and a similarity search, exactly Rekognition's three face APIs, with a crawler that feeds a web-scale collection. Amazon sells the components and leaves the collection to the customer, which is also why there is no opt-out: the faces in any collection belong to whoever built it. The concepts section walks through the full pipeline.
What is the police moratorium?
In June 2020, after the killing of George Floyd and the protests that followed, Amazon announced a one-year halt on police use of Rekognition's face recognition. In May 2021 it extended the halt until further notice without giving a reason or an end date, while continuing to allow anti-trafficking and missing-children organisations to use the service (Euronews, Reuters report, checked November 6, 2026). No lifting of the moratorium had been reported by November 2026. The moratorium covers US police customers; it does not restrict other customers or other countries.
-
2016
Rekognition launches
Image analysis, including face detection and comparison.
-
2018
ACLU test
The ACLU reports that Rekognition matched 28 members of Congress to mugshots at the default threshold; Amazon disputes the threshold used.
-
June 2020
One-year police moratorium
Announced after the George Floyd protests.
-
May 2021
Extended indefinitely
Reuters reports the extension until further notice.
Questions
Can Rekognition find a person on the internet?
No. It searches faces within a collection the customer has built from its own images. There is no web index.
What does Rekognition face search cost?
About USD 1 per 1,000 images at the first pricing tier, plus a small monthly storage charge per stored face vector, as of November 6, 2026.
Can police use Rekognition?
Amazon has barred US police use of its face recognition since June 2020, extended indefinitely in May 2021. Anti-trafficking and missing-children organisations are exempt.
Can I opt out of Rekognition?
There is nothing to opt out of. Faces are stored only in collections that individual customers create, and those customers are responsible for them under the applicable law.
Sources
- 1.AWS: Amazon Rekognition pricing (checked November 6, 2026)
- 2.Euronews (Reuters): Amazon extends moratorium on police use of facial recognition software (checked November 6, 2026)
- 3.The Hill: Amazon extends facial recognition moratorium for police (checked November 6, 2026)
Entry reviewed November 6, 2026. Found an error? Tell us.
Related entries
- Engine Betaface Betaface, the face recognition API and SDK vendor: what the API does, the free demo and daily limit, and how it differs from a web face search.
- Engine Clearview AI Clearview AI: the law-enforcement face search built on scraped web images, its 100-billion-image claim, who may use it, and fines and orders from seven countries.
- Term One-to-many matching (1:N) One-to-many (1:N) face matching defined: a probe face against a gallery of many, how it differs from 1:1 verification, and why errors grow with gallery size.
- Term Similarity score Similarity score defined: the number from comparing two face embeddings, what it does and does not mean, thresholds, and why a high score is a lead, not proof.
- Concept Building a web-scale face index How face search engines crawl the public web, store billions of embeddings for nearest-neighbour lookup, keep the index fresh, and decide what counts as public.
- Concept Accuracy, bias and testing How face recognition accuracy is measured, what NIST FRTE reports found on error rates and demographic differences, and why marketing claims cannot be checked.
Cite this entry
FaceSearch Wiki editors. "Amazon Rekognition". FaceSearch Wiki, updated November 6, 2026. https://facesearch.wiki/engines/amazon-rekognition/