In 45 minutes, Edward Dawson-Taylor from CG Pro shows what virtual production can do right now — from 2D plates and 2.5D layers through to fully rendered 3D environments, alongside newer approaches such as Gaussian splatting and AI-assisted content creation. All of it live on the LED wall with the tracked camera running. Less a lecture than a run through the possibilities: what the technology reaches today, and what that opens up for filmmaking. The aim is that the audience leaves with a real sense of what the medium is now capable of.

In 45 minutes, Edward Dawson-Taylor takes a slower and closer look at how virtual production actually works. Drawing on his work as Virtual Art Department Lead on The Lion King and as a senior artist on The Jungle Book, he goes through the core principles of in-camera VFX — the camera frustum, tracking, the relationship between wall and lens — along with the biggest opportunities the method brings and the challenges that come with it. Worked through on the wall itself rather than on slides. The aim is that the audience leaves understanding how virtual production works, and able to judge where it fits.

Building a Production-grade Gaussian Splatting Pipeline for Virtual Production.

Gaussian splatting is one of the most significant recent breakthroughs in real-time 3D capture. Bringing it into a working film and television production pipeline is still new territory. This session covers the R&D behind a production-grade splatting workflow for virtual production, spanning capture, reconstruction, optimization, and deployment on an LED volume. Laura will share tested methods and results across the pipeline:

• Capture: XGrids, DSLR stills, FX6 video, drone. Objects and sets, indoor and outdoor.
• Reconstruction: structure from motion in Metashape and COLMAP, SLAM.
• Training and optimization: Postshot, Volinga, LichtFeld, and a custom pipeline.
• Editing and cleanup: Postshot, Volinga, SuperSplat.
• Deployment: Unreal via Volinga, then onto the wall.

This is an open look at how we document and optimize the creation and integration of splats, including where the process is already strong and where we are still refining it. It’s exactly as obsessive as it sounds. Audience: Virtual production teams, directors, producers, DPs, technical artists, and filmmakers looking to expand their toolkit.

This workshop is a live experiment. Starting with a blank project and a room full of people, we have 90 minutes to build a 3D environment from a single photograph and display it live on the LED Volume. We will build up a workflow for Gaussian Splatting; nothing is rendered in advance.

Together, we will explore the mechanics of virtual production and its key component: the content. We will show how close we can get to reach a working scenario within the mentioned timecap. By the end, you will have a clear understanding of where and how this technology works today, where it falls apart in front of a camera and what questions your legal department will ask before anything is delivered.

Current media production relies on fragmented information distributed across various information silos maintained by systems and experts, whose terminologies sometimes differ. In addition to metadata, technical information about the media structure, content-related information, and dramaturgical details regarding scene transitions, the intended meaning and interpretation of cinematic elements, as well as artistic standards, play essential roles in media production. As a “digital media twin,” knowledge graphs can integrate this information into a single, authoritative source to be used for a wide variety of processes in media production and consumption.
Following an introduction to knowledge graphs and related concepts, we will use the use case of cataloging film production materials to demonstrate the role that knowledge graphs can play in media production. In this workshop, we aim to explore further potential applications of knowledge graphs.
The workshop will be held by Thomas Hoppe (Fraunhofer Fokus) and Tim Deussen (XRBB) and is presented by GenKI4Media, a project dedicated to the development of generative AI assistants for the media, cultural, and creative sectors and funded by the Federal Ministry of Research, Technology and Space.