UC 1.1 New Media Production

In the media application domain, UC1.1 explores how generative AI can accelerate high-quality 3D asset creation for virtual production. The use case looks at how Multimodal Generalist Foundation Models can give 3D artists, directors and art departments faster, more flexible and cost-efficient workflows, without sacrificing creative control or visual quality.

UC1.1 is part of ELLIOT’s effort to tackle the very concrete pain points faced by today’s virtual production teams.

• Slow and costly 3D asset creation: 3D artists and virtual production crews need faster, cheaper ways to produce high-quality 3D assets. Current workflows rely on specialist tools and manual modelling, which creates bottlenecks when productions need many variations or late changes.

• AI-generated 3D that isn’t production-ready: Early AI tools can generate meshes, but results often lack the fidelity, clean topology and animation-readiness needed for direct use in film, TV and immersive pipelines. Integration with established virtual production setups is still immature.

• Lack of control, consistency and style: Creative teams need intuitive interfaces to steer and refine AI outputs, gaining control over artistic style, geometry and textures, and the ability to generate assets that fit a recognisable house style and brand identity across projects. Today’s tools rarely remember studio-specific preferences or enforce consistent visual rules.

• Workflow fit, IP and privacy concerns: Any new solution must slot smoothly into existing pipelines with easy import/export, while ensuring that sketches, images and generated assets remain private and compliant with IP policies and the EU AI Act. This combination of workflow, legal and trust requirements is not yet met by off-the-shelf tools.

To respond to these challenges, UC1.1 develops an AI-powered copilot for 3D asset creation, built on ELLIOT’s multimodal foundation models and designed to fit real virtual production workflows.

AI assistant for 2D–3D asset creation

• ELLIOT explores generative models that turn text prompts, sketches or reference images into clean, textured 3D meshes.

• The focus is on production-ready quality, so assets can be animated, lit and rendered with minimal manual clean-up.

• Artists stay in control, using the tool to iterate faster on ideas and variations, rather than replacing their expertise.

Built for real-world pipelines

• The solution is conceived as a plugin or companion tool that connects to standard Digital Content Creation (DCC) tools and real-time engines.

• Import/export, preview and versioning follow existing virtual production workflows, so teams do not need to redesign their processes.

• Style controls help studios keep a recognisable, consistent visual identity across projects.

Trust, IP and European values

• ELLIOT’s models are developed with data governance, IP protection and EU AI Act principles in mind, keeping proprietary inputs and assets under the control of creators and studios.

• By combining strong generative capabilities with transparent, controllable interfaces, UC1.1 aims to deliver an AI assistant that is useful, reliable and aligned with European public-service values in media production.

• Media and entertainment companies: broadcasters and studios interested in faster, more flexible production and new formats.

• Creative and technical production teams: 3D artists, modellers, virtual production crews, directors, producers and art departments who feel the day-to-day bottlenecks in asset creation and content delivery.

• Technology and research partners: AI researchers, tool providers and platform developers working on multimodal models, virtual production tools and immersive environments.

• Audiences: in particular younger, digitally native users who will ultimately experience the new formats.