2026 Accelerator Project

Objectives

Champions: DAZN, EBU, IET, RAI, Saudi Broadcast Authority (SBA), SMPTE

Particiants: Eluvio, Google, NeoViews, V-Nova

At IBC2026, the Delta Protocol: Live Media Reinvented Accelerator demonstrated a fundamental shift in live-video delivery: from transmitting every pixel, frame by frame, to transmitting an understanding of what matters in the scene. 

Broadcasters, technology providers and industry bodies came together to test a bold question. Could a live stream carry a lightweight, synchronised layer of scene intelligence that lets the experience adapt on the device, rather than multiplying encoded video variants? In a matter of months, the consortium took that idea from whiteboard to a functioning end-to-end proof of concept.

Innovation 

Live streaming is one of broadcasting's great engineering achievements, but its brute-force model is under growing pressure. Rising compute and CDN costs, device fragmentation, latency expectations and the overhead of delivering differentiated experiences at scale are all pushing against it. 

Delta Protocol treated a live event as a sequence of meaningful moments rather than a stream of pixels. AI models analysed the broadcast in real time. They generated compact "delta" instructions describing what was happening, such as player context, crowd response, weather conditions and key moments. Each instruction was synchronised frame-accurately to the video. 

On the device, those instructions drove the experience. Overlays, enhancements and personalised elements were composed locally, without re-encoding a new stream for every variation. AI stopped sitting around the stream and became part of the delivery architecture itself. 

Industry Collaboration 

Delta Protocol connected capabilities that already exist into one broadcast-grade, interoperable workflow: computer vision, real-time AI, GPU compute, advanced video processing, modern transport and capable client devices. 

Each organisation brought a critical piece: 

• DAZN conceived Delta Protocol and defined its end-to-end architecture 
• V-Nova was evaluated as an alternative feed source for AI analysis. 
• NeoViews and Google Cloud delivered richer scene understanding. 
• Eluvio kept the video and intelligence tracks in sync. 

Alongside champions from across broadcasting and standards bodies, the consortium tested the model in the open and at pace. 

Proof of Concept Demonstrated at IBC: 

The IBC2026 demonstration used as-live football to show a reference end-to-end implementation: 

• A video feed was ingested and analysed in real time. 
• Scene intelligence was converted into compact delta instructions. 
• The intelligence track was transported in sync with the video. 
• A command-centre view visualised the protocol at work across the workflow. 
• Two client devices each received an experience composed for them from the same event. 

It proved the hardest part of the concept works: AI understanding a live sport stream as it unfolds and turning that understanding into instructions a device can act on instantly. 

Key Learnings: 

The Accelerator validated both the promise of AI-native delivery and the engineering discipline it demands. 

Spotting pixel-level change is the easy part, because compression already does it. The real challenge was deciding which changes matter: 

• A player jogging is noise. 
• A goal is signal. 
• A crowd erupting is context. 
• A sudden rainstorm changes what the device needs to render. 

Tuning that threshold on real sports content, at real speed, taught the team more than lab data ever could. 

Three principles emerged: 

• Freshness beats perfection. A useful instruction now is worth more than a perfect one too late. 
• Cloud and device are one system. Understand the scene where it is most efficient, transport only what matters, and let the device render what it does best. 
• Latency is the budget. Every model, decision and data flow must earn its place within milliseconds.  

Industry Impact: 

Delta Protocol established a working reference architecture and a credible direction for the next generation of live media. The approach could enable: 

• hyper-personalised live experiences, composed on the device; 
• lower transport, storage and re-encoding costs by removing duplicated streams; 
• simpler, more efficient processing across the media supply chain; and 
• new approaches to rights protection, watermarking and content integrity.  

The long-term vision goes further. As on-device AI matures, the device could reconstruct the viewing experience from the intelligence track itself. That would make intelligence, not pixels, the foundation of live delivery. 

For DAZN and the wider consortium, IBC2026 marked the completion of the Accelerator and the start of the next challenge: developing, standardising and scaling AI-native transmission for the future of live media.