What it does
Sightengine is a set of APIs for analysing user-submitted media. Moderation covers images, video and live streams, text, audio profanity, and text or QR codes inside images, with 150+ moderation classes for areas such as nudity, hate, violence, drugs, weapons and self-harm. A second group of models flags AI-generated content: images, video, speech and music, and deepfake or face-swap manipulation. For speech and music the site says no watermark or metadata is needed. Other products include face liveness and recapture detection for onboarding and KYC checks, profile image validation, age group estimation, near-duplicate image and video search, OCR, image quality assessment and image or video anonymisation. The site says results come back instantly, the pipeline can scale to tens of millions of images a month, and no human moderator sees the images, which are not shared with third parties. It offers SDKs, documentation and a demo.
Capabilities
- Moderation of images, video, text, audio and text inside images
- Detection of AI-generated images, video, speech and music
- Deepfake and face-swap detection
- Face liveness, recapture detection and profile image validation
- Duplicate and near-duplicate detection for images and videos
- Image quality assessment and OCR
When to use it
- Screening user uploads on a marketplace or social platform for unsafe content
- Checking whether submitted images, video, voices or music are AI-generated
- Confirming a real person is present during onboarding or account recovery
- Blocking spam, duplicates and attempts to share contact details
