Eurovision Sport Deploys Live AI Subtitling at Scale

This partnership between Eurovision Sport and CAMB.AI builds on a collaboration that started in 2024. That year, the two organizations delivered what they described as Europe's first AI-powered, real-time translated sports commentary for European Athletics events. CAMB.AI's technology supports transcription and translation in over 150 languages. The system can output subtitle files in various formats like SRT and VTT, compatible with most broadcast and streaming platforms. Beyond subtitling, the platform is also used for AI-powered dubbing and voice cloning. The implementation for the Milano Cortina 2026 Paralympic Winter Games will cover all six sports, providing real-time, contextual speech-to-text transcription for both live and on-demand content. This initiative is part of the European Broadcasting Union's most extensive digital coverage of a Paralympic Winter Games to date. The technical challenge in live AI subtitling is balancing accuracy with low latency. AI models achieve higher accuracy when they can process more contextual information from the audio, which naturally introduces a slight delay. Some systems use the inherent latency in HTTP Live Streaming (HLS) protocols to improve the quality of automatically generated captions. For newsrooms scaling their own video processing, infrastructure often involves a pipeline architecture. This can include microservices for each stage—like decoding, AI inference, and encoding—which can be scaled independently using orchestration tools like Kubernetes. This modular approach allows for more efficient use of resources, especially costly GPU nodes for AI tasks.