Pangram Labs Releases AI Image and Video Detection Models
Pangram Labs has officially launched Pangram Image 1.0, an AI image detection model that uses heatmaps to locate artificial content, alongside the Pangram Image Flash 1.0 video detection model. The company reported a 99.96% accuracy rate for the image model on its internal test dataset and 98.7% accuracy for the video model on the DF26 dataset. These tools address the growing challenge of identifying synthetic media and deepfakes across the digital ecosystem. Providing robust detection mechanisms helps combat misinformation, digital fraud, and unauthorized manipulation of visual content. Pangram Image 1.0 maintains strong resistance against image compression, resizing, and manual retouching, achieving a 96.4% accuracy rate even at a low resolution of 256x256 pixels. The release also includes features such as native video uploads, Chrome extension integration, and API support for both images and videos.
## BACKGROUND
Deepfakes and AI-generated images are synthetic media created using advanced machine learning techniques, such as generative adversarial networks, which can convincingly mimic real people or objects. Image forensics and detection models use computer vision to analyze visual artifacts and inconsistencies to determine whether media was artificially generated.