Project description
The project centers on training machine learning networks to assist artists and game developers in enhancing rendering performance and quality. The team creates a diverse array of photorealistic 3D human characters at film and cinematic quality, rendered using offline path tracers such as Arnold, RenderMan, and Cycles, and adapts them for real-time engines when necessary. The work encompasses the entire cinematic character pipeline, including sculpting, topology, texturing, and look development of skin, hair, and cloth, as well as facial expressions, grooming, and rigging. The project also engages with emerging rendering standards like MaterialX, OpenPBR, and UE5 Substrate. The training process covers a broad range of complex scenarios involving advanced shading, lighting, motion, post-processing, upscaling, and antialiasing. Each scenario requires high-quality ground truth data that accurately meets production standards for cinematic CGI and game development. Artists involved in the project utilize a variety of modern, industry-standard technologies and software. The primary rendering engines include Blender Cycles, Unreal Engine, and custom renderers based on GLTF, USD, and MaterialX data.
Responsibilities
SKILLS
Must have
Nice to have
- Unreal Engine 5: character setup, Groom, Path Tracer, Substrate materials - Game-ready character optimization (LODs, card-based hair, real-time shading) - Reallusion Character Creator (CC5) and/or MetaHuman workflows - Rigging and animation: skeleton edits, UE5 retargeting, mocap and motion matching data - MaterialX and OpenPBR standards