← ArchiveTalk2023
Motion Mastery: 4D Facial Performance Capture, KineFX and PDG-Powered MoCap Pipeline
3SISTERS studio presents a 4D volumetric capture pipeline for creating Sharon Stone's digital double, coupled with a PDG and KineFX-driven mocap retargeting workflow in Houdini.
Abstract
3SISTERS presents two character pipelines built around Houdini KinFX and PDG. The first covers a digital double of Sharon Stone using photometric facial scans and 4D performance capture at 60 scans per second (over 80TB of raw data), with topology made consistent across frames via proprietary Faceform Wrap topotransfer, plus three-point face stabilization, separate eye and lower-teeth tracking, a Python shot importer, and Houdini-based grooming for hair, brows, eyelashes and peach fuzz. The second walks through an Unreal Engine cinematic using heavily modified MetaHumans, with Xsens-recorded body animation processed and procedurally retargeted in a custom KinFX/PDG tool that filters takes, sets ranges and exports at scale. Prop and instrument rigging uses auto-skeleton aim setups, corrective joint blendshapes fix folds at the hands and knees, cloth is simulated in Vellum with Marvelous Designer material-attribute splitting and PDG-driven film constraints, and JSON metadata carries Houdini layout transforms into Unreal level sequences.
How to read this
- Category
- Production talk: 4D capture and a KineFX/PDG mocap pipeline
- Contributions
- Demonstrates a digital-double pipeline using photometric facial scans and 4D performance capture, with cross-frame topology made consistent via Faceform Wrap topotransfer plus face stabilization and separate eye/teeth tracking
- Shows a procedural KineFX/PDG mocap retargeting tool for Xsens body data that filters takes, sets ranges, and exports at scale for an Unreal cinematic using modified MetaHumans
- Covers prop/instrument rigging with auto aim setups, corrective joint blendshapes for hand/knee folds, Vellum cloth from Marvelous Designer, and JSON-carried layout metadata into Unreal
- Context
- Relates to anatomically-constrained facial performance retargeting (Chandran et al.), applied within a Houdini KineFX and PDG production setting.Builds on: Local Anatomically-Constrained Facial Performance Retargeting
- Correctness
- Studio practice, not peer-reviewed; results are production-proven on specific shows (a Sharon Stone digital double and an Unreal cinematic) at large data scales, so the workflow reflects those projects rather than a general benchmark.
- Clarity
- Pipeline-heavy and tool-specific; a single pass conveys the two workflows, rewatch for the topotransfer and PDG retargeting specifics.
- How to read it
- Watch once for the overall 4D-capture and retargeting pipeline shape, then revisit the topology-consistency and KineFX/PDG retargeting segments if you work in Houdini-to-Unreal character pipelines.
Built upon by
Nothing yet.
Related work
- MetaHuman Framework and Machine Learning for Next-Gen Character Deformation 2023 / GDC
- Creatures in Houdini | Ahmed Gharraph | FMX 2019 2019 / FMX
- Simulating the Perfect Groom for a Bovine Biker | Untold Studios | FMX HIVE 2023 2023 / FMX
- Grooming and Simulation Methods for Different Hair Types | Andriy Bilichenko | Paris HIVE 2023 2023 / Houdini
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →