FRAMES: Federated Retrieval, Agentic Media Environment and Software (Defined Workflows)
Objectives
Champions: EBU, ITV, ETC/USC, XRBB, BBC, RAI, MovieLabs, INSA National Institute of Applied Science, The Global Creative & Security Community
Participants: AMD/HP, Enigmatic, Streamline Media, Eddie AI, AIMICI
FRAMES showed how an ontology-driven, agentic production framework can transform audiovisual archives from static repositories into active creative resources, while maintaining transparency, provenance, and human editorial control throughout the production process.
The Challenge
A production usually generates volumes of information, but they are typically scattered across different systems, departments, and workflows.
It is difficult to discover archive material because metadata is incomplete or difficult to search semantically.
Bridging creative intent and available material is a challenge because AI agents have to understand both narrative requirements and archive metadata.
It becomes difficult to maintain transparency and provenance when multiple AI models and tools are used throughout production.
Achieving photorealism and maintaining control over lighting, proportions, and physical realism remained significant challenge.
The core challenge objective: To be able to use a knowledge graph and a common semantic model to increase visibility into the production process, allowing clearer and better tradeoffs between creativity, time, transparency, and budget.
POC Results
The proof of concept successfully demonstrated an end-to-end agentic workflow for archive-driven content creation through the production of the short film One Small Crawl.
An AI-driven pipeline processed archive metadata and script content to generate shot proposals and production briefs. The system automatically produced a complete shot list in MovieLabs Ontology for Media Creations-compliant formats with archive recommendations linked to narrative requirements. Where archive material was unavailable, generation briefs were used for synthetic content.
The project demonstrated how AI agents can interact with production systems through OMC-compliant workflows. The production assistant integrated documentation, meeting transcripts, workflow metadata, and project tracking systems into a structured knowledge base capable of controlling tools and assisting production activities.
The team successfully created a synthetic protagonist using traditional 3D modelling and rigging and LoRA training based on multiple datasets.
The use of 3D blocking and controlled character assets improved consistency and creative control across generated footage.
One of the most significant achievements was the implementation of a provenance framework capable of recording models and agents used and human revision chains. This created a transparent and auditable production workflow consistent with emerging industry requirements for trustworthy AI use.
Key Learnings
The project highlighted several important conclusions:
- Structured semantic metadata and the MovieLabs OMC framework are fundamental enablers of agentic media workflows.
- Human creative supervision remains essential.
- Effective asset and data management significantly improves production efficiency.
- Archive metadata to dramatically increase historical authenticity and reduce bias.
- AI-generated media quality still requires substantial post-production refinement.
- Transparency, documentation, and provenance tracking must be embedded throughout the workflow rather than added afterwards.
POC Results Showed at IBC2026
Live demos and tutorials were showcased at the IBC Accelerator Zone, and a full-fledged presentation of the POC results, creative process and technical workflows were presented on stage during their Final Showcase Session.
Summary
The FRAMES proof of concept successfully demonstrated that ontology-driven, agent-based production workflows can connect archives, metadata, creative processes, AI generation, and post-production into a unified environment. The project proved the feasibility of automatically transforming archive collections into actionable creative assets, generating structured production plans, and maintaining complete traceability of the resulting media while preserving human oversight
What’s Next?
Several Champions and Participants will continue this R&D to improve the short film and present to conferences, and via new commercial partnerships formed out of this 2026 Challenge.
)
){kind=logo}
)
){kind=logo}
){kind=logo}
)
){kind=logo}
){kind=logo}
){kind=logo}
){kind=logo}
){kind=logo}
){kind=logo}
){kind=logo}
){kind=logo}