With DETECT-World now on the platform, deepfake detection plugs into the existing workflow without adding another account or integration to manage. Most detection systems recognize the statistical artifacts that specific generators leave behind, which means they can struggle when a brand-new generator launches. DETECT-World adds a second layer. Alongside artifact scanning, it checks whether lighting, motion, and spatial consistency in a scene actually make sense. That physical-plausibility check helps it detect content from generators it was never trained on. The model has been benchmarked against more than 250 generative models across 54 languages. On audio, it scored 99.47% accuracy on a third-party Podonos benchmark, ranking first among 18 systems tested. On image and video, internal benchmarks put accuracy at 95.77% and 98.2% respectively. 

Source: https://www.resemble.ai/resources/detect-world-eden-ai-deepfake-detection-api