What is MXL?
Media eXchange Layer (MXL) is an open-source project for real-time, zero-latency, vendor-agnostic exchange of video, audio, and timed metadata between processes in software-based broadcast environments. Just as SDI allows any device to send uncompressed, almost latency-free video to any other device with a simple cable, MXL brings the same to media in software processes. Developed under the Linux Foundation as part of the European Broadcasting Union's Dynamic Media Facility (DMF) architecture, MXL solves a fundamental challenge of creating vendor-agnostic, uncompressed media workflows in software, which aims to remove complexity and vendor lock-in.
Why MXL Matters
While file-based workflows have been software-centric for a long time, for live video, the broadcast industry is still in the process of moving from dedicated hardware to flexible, software-based infrastructure. Many vendors offer software that either runs fully in the cloud or on a server-as-an-appliance for on-prem use. The logical next step in this process is to be able to allow software to run on generic servers on-prem as well as in-cloud, providing hybrid workflows. Dynamically running software from different vendors on the same hardware or on adjacent hardware promises benefits in flexibility, scaling, efficiency and vendor choice.
However, without MXL, every vendor is forced to implement media exchange differently. Traditional IP protocols like ST 2110 were designed for hardware-to-hardware connections via the network and become complex and inefficient when used between software applications on the same server or compute cluster. MXL provides a vendor-neutral, open standard specifically designed for software-to-software uncompressed media exchange, restoring the freedom of choice that SDI provided in the hardware era.
How MXL Works
MXL uses shared memory to move media between applications: instead of packaging video and audio into network packets that travel through multiple software and hardware layers, applications simply write their output to a designated area of computer memory whilst other applications read directly from that same space. This approach minimises overheads, with applications communicating through standard MXL APIs regardless of vendor. The result is near-instantaneous uncompressed media transfer with latencies below one millisecond per hop, compared to at least 20 milliseconds per device with ST 2110, meaning you can chain together multiple processing steps without accumulating problematic delays.
Current Implementation and Next Steps
While MXL continues to use shared memory in a single server, it will never address the needs of a larger organisation, which is exactly the type of company most likely to benefit from MXL. This is why the next phase of development is focusing on extending MXL across multiple servers using RDMA (Remote Direct Memory Access) technology, which would allow one machine to access another's memory across an IP network whilst bypassing the CPU. This would be key to allowing MXL-based workflows to scale across distributed compute clusters and cloud environments whilst maintaining low-latency characteristics to deliver large-scale workflows.
Use Cases in Live Production
Major sporting events like the Olympic Games and other world-famous events often require 3–4 times peak capacity compared to normal operations. As part of a wider software infrastructure, MXL allows broadcasters to stage/test infrastructure before even getting on site. At games time, they can then dynamically spin up additional replay servers, clipping services, and multiviewer instances as needed, retiring them after the event. Moreover, MXL enables new ways to build best-of-breed production workflows.
MXL enables asynchronous "faster-than-live" workflows, which reduces latency and increases efficiency. For sports broadcasters managing multi-format delivery across linear, OTT, apps, and social platforms, the framework allows integration of specialist tools like AI-based highlights and best-in-class tracking without building custom integration code for each. The vendor-agnostic design means production workflows can mix media functions from multiple manufacturers without bespoke integration projects, reducing both complexity and vendor lock-in while enabling broadcasters to select best-of-breed tools for each production element.
Built Differently
The MXL project involves unusually broad, or at least unusually open, industry collaboration and is publicly available on GitHub. Started in April 2025 by the Linux Foundation, European Broadcasting Union, and North American Broadcasters Association, the initiative includes major broadcasters like the BBC, CBC/Radio-Canada, France Télévisions, SVT, RTÉ, and VRT actively contributing to development and testing in real production scenarios. We are working alongside technology vendors including AWS, NVIDIA, Grass Valley, Intel, Lawo, and Riedel alongside the broadcasters to make sure MXL meets real-world requirements. This approach is key to making MXL a global, vendor-neutral standard, with challenges solved in the open rather than independently by each vendor.
The Road Ahead
Early demonstrations at IBC 2025 have shown MXL working effectively, with software from different vendors exchanging live media seamlessly. In 2026, both NAB and IBC will see MXL products becoming available as well as some broadcasters starting trials. There is work within the EBU and AMWA to complete the timing and synchronisation specifications as well as part of the Dynamic Media Facility Joint Taskforce (JT-DMF).
As the broadcast industry continues its transition to software-defined production, MXL provides a foundation for building flexible, scalable, and interoperable systems, but will just be one part of a much wider raft of technology coming together to harness the benefits of IT whilst striving to maintain the reliability, robustness and quality of broadcast services we've come to love over the past century.