AWS Launches Edge AI Tool for Real-Time Video Transformation
Amazon Web Services has launched a new tool that uses edge AI to convert horizontal video streams into vertical formats in real time. The technology, which performs low-latency cropping and adaptation on edge devices, is aimed at live broadcasters but has applications in retail and logistics for visual analytics. Fox and NBCUniversal are among the initial customers.
- The service, officially named AWS Elemental Inference, integrates directly into the AWS Elemental MediaLive console, allowing video operators to enable the feature without re-architecting their existing production workflows. - The AI operates with a latency of just 6-10 seconds from live stream to vertical output, a significant improvement over traditional post-production methods that can take several minutes. - The technology functions as an "agentic AI," meaning it analyzes video feeds and makes cropping and clipping decisions autonomously without the need for human-in-the-loop prompting or operation. - Fox Sports collaborated with AWS for 18 months to train a custom machine learning model that understands the specific editorial and creative standards for its sports content. - The system creates a "saliency map" of the action on screen, which goes beyond simple object tracking to balance object detection with contextual awareness, ensuring the reframed shot is coherent. - In addition to live-to-vertical conversion, the tool's advanced metadata analysis can automatically detect and generate highlight clips of key moments like goals or game-changing plays for immediate distribution. - Fox Sports is already using AWS Elemental Inference to create vertical highlights for LIV Golf events and intends to use the technology for all 104 matches of the upcoming FIFA World Cup. - In beta testing, large media companies using the AI-powered workflow achieved cost savings of 34% or more compared to using multiple, separate video processing tools.
