Mo-Sys Engineering, a leader in virtual production innovation, is experimenting with Gaussian Splat, a revolutionary scene creation technique that has the potential to transform the way photorealistic content is produced.
Virtual production is entering an exciting new phase, where the creation of production-ready photorealistic 3D content is being radically simplified.
Previously photogrammetry looked to have solved this problem but was ultimately found to be too complex. Then came NeRF, a smarter way of generating 3D environments from video footage with the help of AI. This technique did simplify the capture process but introduced impractical render delays.
Gaussian Splat is enabling a new, simple and rapid scene capture process without AI, and the performance sapping issues associated with NeRF.
Born out of a European Research Council funded project and pioneered by Bernhard Kerbl and Georgios Kopansas of the University Cote D’Azur, Gaussian Splat is making it possible to create photorealistic scenes quickly from a variety of sources, including drone footage, handheld cameras, and even smartphones.
Mo-Sys is now experimenting with Gaussian Splat and has already found several ways of implementing it for use inside Unreal Engine. This is the beginning of an exciting journey and Mo-Sys is committed to sharing details of our learning.