Dalet is bringing NVIDIA's Synthetic Video Detector into Dalet Flex so media organisations can analyse incoming images and video for signs of AI generation or manipulation. Instead of presenting the result in a standalone detection tool, Flex records the signal as asset metadata and can use it to trigger notifications, approval stages or an additional human review.
The models can examine images, individual frames and video content for patterns associated with diffusion-generated material. Dalet says analysis can run where the media resides, including on-premises and hybrid environments, through Dalet Flex and Dalet AI services. That positioning is particularly relevant to newsrooms, where potentially synthetic material must be evaluated without breaking the chain of custody or moving sensitive content into an uncontrolled service.
The key technical contribution is the link between probabilistic detection and operational action. A confidence signal only helps a newsroom when it travels with the asset and feeds a governed process that can stop publication pending verification. Dalet explicitly keeps human validation in the loop. The capability is currently offered for proof-of-concept deployments, so production accuracy, performance on compressed or edited material and the operational thresholds used by customers remain open questions.
