Amazon Rekognition Content Moderation
aws.amazon.com
- Free plan
- Web
- API
What Amazon Rekognition Content Moderation is
Amazon Rekognition Content Moderation is a cloud service that uses machine learning and managed APIs to moderate images and video. It identifies categories such as explicit or suggestive material, violence, drugs, tobacco, alcohol, hate symbols, gambling, and disturbing content. Results include labels and confidence scores; video results also include timestamps. Teams can use a hierarchical taxonomy to set rules at different levels of detail. For image analysis tailored to a particular use case, customers can train a Custom Moderation adapter with as few as 20 annotated images. Low-confidence predictions can be sent to Amazon Augmented AI for review by a team or third-party reviewers. Stored-video analysis can use an Amazon S3 bucket and publish its completion status to an Amazon SNS topic. The service supports image and video moderation, custom policies, human review, and cloud deployment. AWS says billing is based on images or video duration processed, with no minimum fees or upfront commitments. A free tier lasts 12 months from account creation and includes 60 video-analysis minutes and two adapter-training hours per month.
Who it suits
It suits teams moderating images or video for social media, gaming, ecommerce, advertising, and media and entertainment. Organizations that need tailored image rules or human review can use custom adapters and Amazon Augmented AI.
Good to know, and what to watch
- Detects a range of image and video content categories
- Video results include timestamps and confidence scores
- Custom adapter training starts with 20 annotated images
- Low-confidence results can go to human reviewers
- Free tier includes monthly video minutes and training hours
- Free tier lasts 12 months from account creation
- Images and videos may be stored unless customers opt out
- No text or audio moderation
Plans
Every plan and priceCustom Moderation adapter
Standard image moderation (DetectModerationLabels, Group 2)
Video content moderation
Amazon Rekognition Custom Moderation adapter
The verdict
Amazon Rekognition Content Moderation combines category detection with configurable rules, custom image models, and a human-review option. Its data-use default and time-limited free tier are worth considering before adoption.
What the maker publishes
- API pagination
- GetContentModeration supports MaxResults and NextToken pagination when a video produces many moderation labels. source
- API workflow
- For stored video, StartContentModeration analyzes a video in an Amazon S3 bucket and publishes completion status to an Amazon SNS topic. source
- Company founded
- AWS states it launched in 2006. source
- Company headquarters
- LinkedIn lists Seattle, Washington as AWS headquarters. source
- Company legal name
- The maker is Amazon Web Services (AWS), the cloud business of Amazon.com, Inc. source
- Company size
- LinkedIn lists AWS company size as 10,001+ employees. source
- Compliance
- Third-party auditors assess Amazon Rekognition as part of multiple AWS compliance programs, while customers remain responsible for evaluating their own compliance obligations. source
- Compliance
- AWS states that third-party audits cover Rekognition under SOC, PCI, FedRAMP, HIPAA, and other compliance programs. source
- Confidence control
- Developers can set MinConfidence to trade recall for precision when filtering moderation results. source
- Custom models
- Amazon Rekognition Custom Moderation lets customers train an adapter with as few as 20 annotated images for image analysis tasks tailored to their use case. source
- Custom models
- Custom Moderation trains an adapter on annotated images and supplies its AdapterID to DetectModerationLabels to improve results for a specific use case. source
- Data handling
- Images and videos passed to Rekognition may be stored and used to improve the service unless the customer opts out; the stored content is encrypted at rest using AWS KMS. source
- Deployment model
- TrustRadius classifies Amazon Rekognition as SaaS and reports no mobile application. source
- Detected categories
- The API detects adult content, violence, weapons, disturbing imagery, drugs, alcohol, tobacco, hate symbols, gambling, and rude gestures. source
- Detection results
- The service returns detected labels and video timestamps with confidence scores, which can be used with a hierarchical taxonomy to create granular rules. source
- Documentation depth
- AWS documentation covers getting started, SDK installation, images, stored video, streaming video, moderation APIs, custom adapters, and human review workflows. source
- Founded
- 2006 source
- Headquarters
- Seattle, Washington, United States source
- Hierarchical labels
- Moderation results use a three-level taxonomy with L1, L2, and L3 labels and include each label's taxonomy level. source
- Human review
- The service integrates with Amazon Augmented AI so teams or third-party reviewers can assess low-confidence predictions. source
- Human review integration
- Amazon Augmented AI is integrated so human reviewers can handle a smaller set of machine-flagged content. source
- Image API
- Images can be moderated synchronously with DetectModerationLabels or asynchronously with StartMediaAnalysisJob and GetMediaAnalysisJob. source
- Input requirements
- The documented image procedure accepts JPEG or PNG input supplied as bytes or as an Amazon S3 object. source
- Integrations
- The product page describes using Amazon S3 to store uploaded content and Amazon Augmented AI for human review. source
Checked against the maker's pages on 15 Oct 2026.