Summary
NVIDIA is expanding its collaboration with the European Broadcasting Union (EBU) to help public service broadcasters adopt AI while maintaining control over their data, models and infrastructure. At the same time, Adobe is using NVIDIA’s accelerated computing technology to improve video search in Frame.io, helping production companies such as Germany’s UFA find specific scenes in large video archives more quickly.
The developments highlight how AI is being used to address two important media industry challenges: protecting sensitive editorial information and making large collections of video content easier to manage.
What Happened?
NVIDIA announced advancements in AI-powered media workflows at a creative industry event held in Berlin in collaboration with Adobe, Bertelsmann, NVIDIA and UFA.
The announcements focus on two areas. The first involves expanding sovereign AI standards for European public service broadcasters through the EBU. The second involves improving video search and media processing through Adobe Frame.io and NVIDIA’s GPU technology.
Both initiatives aim to help media organizations use AI more effectively without losing control over important content, data and production processes.
NVIDIA and EBU Expand Sovereign AI Standards
The EBU and NVIDIA are expanding their existing initiative to help European public service media organizations adopt AI under standards that prioritize data sovereignty, security and editorial independence.
The expanded framework brings together sovereign cloud providers, developers of open AI models and systems integrators. The goal is to establish practical requirements for deploying AI agents across broadcasting operations.
AI agents could support activities such as newsroom workflows, media production and audience services. However, broadcasters must ensure that sensitive information remains protected and that AI systems operate within their organizational and editorial policies.
The initiative will define reference architectures and requirements covering where AI workloads run, how data is governed, which models are used and how security is maintained throughout a system’s lifecycle.
According to the announcement, the EBU represents more than 115 member organizations across over 55 countries, reaching more than one billion people.
Microsoft, Nebius and Open Models Support the Initiative
Several technology providers and AI model developers are expected to contribute to the expanded framework.
- Microsoft: Its sovereign cloud capabilities, including Azure Local and Foundry Local, can support local processing of sensitive editorial material and help critical services continue operating when external connectivity is unavailable.
- Nebius: Its platform can help broadcasters train, fine-tune and run AI models on European infrastructure.
- NVIDIA Nemotron: These open models can provide a foundation for customized AI applications tailored to local languages, broadcasting archives and public service requirements.
- Qvest and other system integrators: These companies can help translate the proposed reference architectures into practical solutions for real-world broadcasting environments.
Open models can give media organizations greater visibility into how their AI systems work and more flexibility to adapt them to specific needs.
The initiative is intended to help broadcasters introduce AI while maintaining control over sensitive data, infrastructure and editorial processes.
Adobe Frame.io Uses NVIDIA GPUs to Improve Video Search
The second announcement focuses on video production.
Adobe is using NVIDIA accelerated computing technologies to improve how creative teams search, process and manage video content through Frame.io.
Large production companies often maintain extensive video archives containing years of footage. Finding a particular scene, facial expression or conversation can take considerable time when editors have to search manually through files and folders.
Frame.io’s semantic search capabilities are designed to help users locate relevant moments based on concepts, actions and settings rather than relying only on filenames.
The workflow uses several NVIDIA technologies:
- NVIDIA NVDEC: Accelerates video decoding.
- NVIDIA TensorRT: Optimizes AI inference for faster processing.
- CUDA: Supports preprocessing and the generation of AI embeddings used in search.
- NVIDIA NVENC: Accelerates video encoding.
Adobe is also improving media ingestion workflows to help production teams process growing video libraries more efficiently.
Reported Performance Improvements
According to the announcement, the updated video search pipeline delivers several performance improvements.
| Performance Metric | Reported Improvement |
|---|---|
| Embedding generation | About 5 times faster |
| Search processing throughput | Up to 4 times higher |
| Search processing latency | Up to 5 times lower |
| New video proxy size | About 80% smaller than the previous 360p version |
These figures are reported results from the announced workflow and should not be interpreted as guaranteed improvements for every system or production environment.
The smaller proxy format is designed to reduce storage and data transfer requirements while retaining enough visual information for search.
How UFA Could Benefit From Faster Video Search
UFA, one of Germany’s major film and television production companies, produces more than 3,500 hours of programming annually, according to the announcement.
Its productions include the long-running daily drama Gute Zeiten, schlechte Zeiten.
For editors working on a recap or flashback, locating a specific reaction or unresolved conversation from an episode produced many years earlier can be difficult. Traditional folder structures and filenames may not provide enough information to identify the exact moment.
Semantic search can help editors find footage by describing what appears in a scene. For example, an editor could search for a character looking concerned, narrow the results by character name and filter for approved footage.
Hybrid search can combine these results with information such as transcripts, production notes and approval status.
This approach could reduce the time spent searching archives and allow editors to spend more time on creative work.
Why These Announcements Matter
The developments illustrate two ways AI could change the media industry.
First, sovereign AI infrastructure may help public service broadcasters introduce advanced AI tools while retaining greater control over where their information is processed and how their systems are managed.
Second, GPU-accelerated search could make archived video more useful by helping production teams discover relevant footage without manually reviewing large amounts of material.
These capabilities could be particularly valuable for broadcasters and studios managing extensive archives, multilingual content and demanding production schedules.
However, organizations will still need to evaluate security, accuracy, infrastructure costs and the suitability of AI-generated search results for their specific workflows.
What Comes Next?
The next stage will involve turning sovereign AI standards into deployable broadcasting solutions and integrating faster search capabilities into practical production workflows.
For broadcasters, adoption will depend on how well these systems meet local data governance requirements, security expectations and editorial policies.
For production companies, the value of AI-powered search will depend on its ability to identify relevant footage accurately, integrate with existing editing tools and reduce the time and resources required to manage archives.
The announcements also reflect a broader collaboration between Adobe and NVIDIA that extends to tools and workflows involving Substance 3D, Firefly Foundry and enterprise digital twins.
Conclusion
NVIDIA’s expanded work with the EBU and Adobe’s improvements to Frame.io demonstrate how AI infrastructure and accelerated computing can address different challenges in media production.
The EBU initiative focuses on helping European broadcasters adopt AI while protecting data sovereignty and editorial independence. Meanwhile, Adobe’s GPU-accelerated video search aims to help production teams find relevant footage faster and manage growing media libraries more efficiently.
As media organizations continue to explore AI, balancing innovation with security, editorial control and practical production needs will remain essential.