MXL as the bridge between AI and broadcast infrastructures

Artificial intelligence is opening valuable new possibilities for broadcast workflows, from automated quality control and compliance analysis to metadata enrichment, content understanding and intelligent monitoring. But connecting AI services to a live broadcast environment raises a fundamental architectural question: how can broadcasters benefit from fast-moving software innovation without compromising the deterministic behaviour, timing and resilience required by their core media infrastructure?

At IBC2026, BBright’s Guillaume Arthuis presented the technical paper “MXL as an interoperability boundary between artificial intelligence and broadcast infrastructures.” The paper explores how the Media eXchange Layer (MXL) can provide a clear, controlled interface between these two technological worlds.

The central idea is simple: an AI provider should not need to become an expert in SMPTE ST 2110, just as a broadcast infrastructure manufacturer should not need to develop expertise in every AI framework and model. MXL can establish the common boundary through which media and related data are exchanged, while each side remains focused on its own responsibilities.

Two worlds with very different requirements

Live broadcast infrastructure is designed around precise and predictable behaviour. Video, audio and metadata must remain synchronised. Latency must be controlled. Systems must operate continuously, including under heavy workloads, and operators need clear supervision and reliable fallback mechanisms.

AI and general-purpose software services often come from a different environment. They may use asynchronous execution, elastic processing, containers, GPU acceleration and rapidly evolving software dependencies. Their execution time may vary according to the model, the workload or the available compute resources.

Neither approach is inherently better; they have simply been designed for different purposes. The engineering challenge is therefore not to force one world to behave exactly like the other. It is to define a robust interface between them.

MXL as an interoperability boundary

MXL offers a way to exchange media between software functions through an open and structured layer. In an AI-assisted broadcast workflow, the broadcast platform can remain responsible for the functions that must stay deterministic: ST 2110 ingest and output, synchronisation, signal compliance, buffering, supervision and fallback.

AI or software-based services can then operate within a controlled extension zone. They can receive the relevant media grains and metadata, perform tasks such as detection, classification or enrichment, and return results without becoming responsible for the behaviour of the complete real-time media path.

This separation creates a useful division of responsibilities:

• the broadcast core protects timing, continuity and operational control;
• MXL provides the media exchange and interoperability boundary;
• software and AI services bring specialised capabilities that can evolve independently.

For broadcasters, this model creates a practical path towards more open and modular infrastructures. New functions can be introduced, updated or replaced without redesigning the entire media chain, provided that orchestration, resource allocation, monitoring and fallback are handled with the same discipline as the core broadcast system.

Openness without sacrificing operational control

Interoperability should not mean giving uncontrolled third-party components authority over the live media path. An MXL-based architecture must preserve a clear separation between media exchange, control and orchestration, and operational fallback.

This is particularly important for AI. A model can deliver powerful results while still presenting variable execution times, changing dependencies or occasional uncertain outputs. In a production environment, those characteristics must be isolated and supervised. The broadcast platform must retain the ability to bypass a service, validate its output or continue operating if that service becomes unavailable.

The objective is therefore not to claim that MXL makes every software component deterministic. Its value lies in defining the boundary at which software innovation can connect to a broadcast-grade system without taking control away from that system.

A foundation for modular, software-defined broadcast systems

The approach discussed in the paper supports a broader evolution towards software-defined media infrastructures. Functions can be deployed as services, isolated through containers and managed according to their own lifecycle, while the underlying platform continues to provide the timing, resilience and observability expected in professional broadcast operations.

This separation also allows specialists on both sides of the interface to contribute effectively. AI developers can work with familiar data structures, frameworks and deployment models. Broadcast manufacturers can continue to concentrate on reliable signal processing, transport and system control. MXL becomes the shared layer that makes collaboration possible without requiring every participant to master the complete technology stack.

Download the paper and presentation

The complete technical paper provides a detailed examination of MXL’s role at the boundary between artificial intelligence and broadcast infrastructures. The presentation offers a concise overview of the main architectural principles and conclusions discussed at IBC2026.

Technical Paper

MXL as an interoperability boundary between artificial intelligence and broadcast infrastructures

Presentation

MXL as an interoperability boundary between artificial intelligence and broadcast infrastructures — IBC2026 presentation

Continuing the conversation

MXL is not intended to replace SMPTE ST 2110 or the discipline required to operate live media systems. It is a complementary technology that can help broadcasters connect their established infrastructure with a growing ecosystem of software and AI services.

For BBright, the strategic direction is clear: openness where the ecosystem needs interoperability, and stability where broadcast operations need trust.

To discuss MXL-ready architectures, AI integration or the evolution of software-defined broadcast workflows, contact the BBright team.

Publication and rights notice

This technical paper was first published as part of the IBC2026 Technical Papers Programme. It is republished by BBright in accordance with the non-exclusive copyright licence agreement entered into with International Broadcasting Convention LLP (IBC). Copyright in the paper and presentation remains with the author(s) and/or their respective rights holders. All rights not expressly granted are reserved.

IBC, IBC2026 and any associated names, logos or marks are the property of International Broadcasting Convention LLP and/or their respective owners. Their use on this page is solely for factual identification of the event and programme. Unless expressly stated otherwise, publication on this website does not imply endorsement of BBright, its products or services by IBC or any of IBC’s owner organisations.

The downloadable presentation is the original author-created presentation. It does not contain, reproduce or provide access to an IBC-owned recording of the conference session.