For those who don’t know the Smpte: The Society of Motion Picture and Television Engineers (SMPTE) sets the technical standards behind nearly everything that moves on a screen. From colorimetry to file formats, SMPTE defines the backbone of professional workflows in postproduction, broadcast, and digital cinema. Their new engineering report Artificial Intelligence and Media (ER 1011:2025) updates the 2023 edition with the latest thinking on AI, and is freely available to non-members.
Open source and open AI
The first section revisits the foundations of machine learning and deep learning, before expanding into what SMPTE now calls “open source artificial intelligence”. The report clarifies the difference between open-weight and truly open source AI, where not just the model weights, but also code, data, and documentation are publicly accessible. It also points to the emerging concept of Responsible AI Licenses and frameworks like the Model Openness Framework (MOF). SMPTE notes that “open-washing”, marketing closed systems as open, is a growing issue engineers must recognise.
Generative models and interoperability
A substantial chapter examines generative AI, from diffusion models and GANs to LLMs and multimodal architectures capable of handling text, sound, image, and video simultaneously. The report highlights the growing influence of protocols like the Model Context Protocol (MCP) and Google’s A2A for agent-to-agent interoperability, described as the foundation for “agentic workflows.” SMPTE sees these protocols as key to connecting intelligent tools in post and broadcast environments, without locking developers into single ecosystems.
Security, ethics, and regulation
Security receives dedicated attention. SMPTE warns of risks such as data poisoning, jailbreaking, and intellectual property leakage within AI-driven workflows. The report advocates Zero Trust Architectures (as defined by MovieLabs), calling for authentication and authorisation mechanisms even between AI agents. The ethics chapter, meanwhile, argues that AI in media must be both transparent and auditable “because failing is expensive, and media needs its own voice.” It outlines principles of inclusivity, transparency, and fit-for-purpose design, forming what SMPTE calls the “AI Ethics Pipeline”.
Standards and data: the next frontier
In its closing chapters, the report maps the AI standards landscape, noting gaps around ontologies, benchmarking, and model metadata. The authors identifie opportunities for collaboration with the EBU, ISO/IEC, and other bodies, stressing the need for shared datasets to train and validate media-specific AI systems. These datasets, the report says, are crucial for reproducible research and safe deployment in production pipelines. As always, SMPTE closes with a reminder that technology, especially AI, must prove itself in real-world conditions before being trusted in production.
The full Engineering Report ER 1011:2025 – Artificial Intelligence and Media is available for free download from the SMPTE website. It is dense, detailed, and refreshingly un-hyped. A technical must-read for anyone designing, coding, or approving AI-assisted tools in post, broadcast, or digital cinema. Before your next machine learning experiment runs in production, take an hour to read what SMPTE’s engineers have already tested and defined.
