View, browse, and sort the ever-growing list of sessions by topic and format.
For more information about conference topics and session formats, visit the GDC Showcase Conference page.
Ashton Mason (Senior Software Engineer, Electronic Arts)
Date: Monday, March 15
Time: 11:50am - 12:20pm
Format: Discussion Session
Vault Recording: TBD
This GDC 2020 talk will feature real-time interaction with the speaker through online chat messaging during its scheduled broadcast. In early 2014, the Frostbite team started developing an in-house LOD technology for quad-based Bioware character models. Today that tech powers tools used by studios across Electronic Arts, from Battlefield to Plants vs Zombies. Quad-based models are widely used in the industry, and artists devote much care to their topology and edge flow. Yet simplification of quad meshes is less studied in the literature -- and popular LOD tools triangulate meshes and ignore topology. While LODs are sometimes thought of as throwaway, maintaining good topology benefits subsequent hand-editing, lighting and animation. For that reason, Bioware historically spent days per model hand-authoring their character LODs. Instead, this work aims to procedurally generate LODs that look like artist-authored models, most of the time, saving hundreds of thousands of dollars per title. A lot was learned, and challenges were overcome. This talk will describe how they were solved.
Attendees will gain insight into the development of a state-of-the-art quad-based procedural LOD generation tool for game development use. They'll see how an academic algorithm was harnessed and improved to meet the needs of real-world meshes, and how a range of interesting smaller challenges were approached and solved.
This talk will appeal to developers and technical artists interested in quad mesh simplification and procedural mesh processing generally. It may also interest artists after in-depth knowledge of procedural LOD generation tools. No programming knowledge is needed and no previous experience of mesh simplification techniques is assumed.