Resemble AI has launched DETECT-3B Omni, a multi-modal deepfake detection model that integrates audio, image, and video detection capabilities into a single API. With 3 billion parameters, this model combines enhanced audio detection from its predecessor, DETECT-2B, with new vision models trained on millions of images and hundreds of thousands of video clips. It addresses the growing threat of synthetic media, which can now span multiple formats in a single attack, making it essential to have a detection system that performs well across all modalities without compromising accuracy. The audio component of DETECT-3B Omni has been significantly improved, featuring expanded training data that includes diverse noise conditions and codecs. It now supports over 40 languages, validated by the MLAADv8 dataset, and can detect outputs from various commercial voice AI tools. Notably, it includes defenses against replay attacks, which are attempts to evade detection by recording synthetic audio through physical devices. In terms of image detection, the model shows over 99% accuracy on images generated by popular tools like StyleGAN and DALL-E, and it can also identify partially edited images, addressing a common threat where genuine images are manipulated. For video detection, DETECT-3B Omni achieves an equal error rate of about 4.5%, with strong performance on recent video generation models. The model's versatility allows for seamless integration into existing systems, requiring minimal adjustments for users of Resemble Detect. Media files can be processed individually or in batches, and users have access to a web-based dashboard for analyzing detection results. Looking ahead, Resemble AI plans to expand video coverage, develop real-time detection capabilities, and enhance defenses against adversarial attacks as the synthetic media landscape continues to evolve.