The Challenge
Broadcasters relied heavily on manual review and editing to generate post-match highlights.
The process was time-consuming, labor-intensive, and delayed content delivery.
Detecting key moments like “six”, “out”, or “goal” needed precise frame detection.
The traditional model training cycle involved manual data labeling, slowing down development.
The Solution
Used AWS MediaLive and MediaReplayEngine for live streaming and frame extraction.
Used SageMaker Ground Truth for automated data labeling.
Built custom Amazon Rekognition models to identify in-game events in real time.
Used AWS Lambda for quick inference and highlight extraction.
Enabled ad insertion during idle gameplay using live state detection.
The Result
30% increase in viewer engagement through real-time replay and highlight distribution.
80% reduction in manual effort, improving editorial efficiency.
25% cost savings via serverless architecture and on-demand inference.
Highlight reels compiled within 30 minutes post-match, accelerating content delivery timelines.
