What it does
Resemble AI's site presents a platform for detecting synthetic media. Its DETECT-World model checks audio, video and images and returns a score and verdict with human-readable explanations. The site cites accuracy of 99.5% for audio, 98.2% for video and 95.8% for images, noting that the audio figure comes from a third-party benchmark while image and video results are internal. Related products include PerTh multimodal watermarks, an Intelligence layer for explainable results, Identity for enrolling and matching voices and likeness, Signal alerts for known fraud patterns, meeting protection that works with Microsoft Teams and other platforms, agent detection, and a Chrome extension. Developers can call detection through a REST API and a Python package. For regulated environments the site lists GDPR, ISO 27001, SOC 2 Type II and HIPAA-aligned data handling, air-gapped on-premises deployment and a zero-retention mode.
Capabilities
- Detects AI-generated audio, video and images with a score, verdict and plain-language explanation
- PerTh imperceptible multimodal watermarks for provenance and C2PA compliance
- Real-time protection for meetings and contact-centre calls
- Voice and likeness enrolment for identity matching
- REST API, Python package and Chrome extension
- Air-gapped on-premises deployment and a zero-retention mode
When to use it
- You need to screen calls, meetings or uploaded media for deepfakes
- You want to add deepfake detection to an app or pipeline through an API
- You want to watermark media to show where it came from
- You need to verify a person's identity during onboarding or KYC
