Software Stack

AI concept generation and 3D base mesh creation from approved 2D concepts.
Runs locally only on controlled GPU hardware, no cloud upload of any production asset.
Returns in post-production for render enhancement and the custom Face Scion LoRA pass.

Works alongside ComfyUI for photobashing, painting, and moodboard production.
AI-generated outputs come into Photoshop for editing, compositing, and final refinement before art direction approval.

PortalCam is a handheld scanner with 96-channel LiDAR and four calibrated cameras that captures real-world environments as 3D Gaussian Splat point clouds up to 60m, with real-time preview on device.
LCC Studio processes the raw captures, cleans the point cloud, optimises splat density, and exports PLY files ready for Volinga import into UE5.

AI-powered 3D Gaussian Splat environment generation from text or reference images.
Used in pre-production for rapid environment exploration and director review before committing to a physical scan.
Where the result holds quality, the Marble asset goes straight to production as a final deliverable, exported as PLY and brought into UE5 via Volinga the same way as any XGrid capture.

UE5 plugin that imports and relights PLY Gaussian Splat data with Lumen-compatible shadow casting and full ACES colour management.
Fully source-agnostic. XGrid scans and Marble environments are treated identically, so real and AI-generated locations can coexist in the same scene with no pipeline changes.

Virtual camera system. The full UE5 scene, characters animated and Gaussian Splat environments loaded, is packaged as USDZ and opened on an iPhone or iPad.
The director navigates with a live handheld camera and records tracked takes with full lens metadata.
Camera paths import into UE5 via Live Link as approved shot references, driving the final render camera in production.

First DCC validation point for the AI-generated base mesh, brought in via a custom GLB import tool.
Hard surface prop and environment hero asset modelling, plus retopology. Not used for animation.

Cloth simulation and costume construction at the pattern level with physically accurate fabric.
Output goes into Maya and ZBrush for retopology and detail, then into Substance Painter for full PBR texturing before final assembly in CC5.

PBR texturing and material authoring, with all maps authored to ACES colour space.
Delivers albedo, roughness, metallic, and normal maps into CC5 for character finalisation and directly into UE5 for environment assets.

Central render hub where all departments converge.
Lumen GI and reflections, Nanite virtualised geometry, Sequencer for shot assembly and AOV EXR output per layer.
Live Link receives iClone animation and Jetset camera data in real time.

Procedural world-building plugin inside UE5. Surface scatter, terrain generation, procedural mesh placement for roads, cables, and organic forms.
Replaces complex PCG graph workflows with an artist-friendly interface, so environment layout is built and iterated without ever leaving UE5.

Neural VDB compression plugin for UE5, achieving up to 100x compression with negligible quality loss.
Custom volumetric renderer runs up to 2x faster than Unreal's native Heterogeneous Volumes.
Receives VDB output from both JangaFX tools and Houdini. Available via Epic Fab.

Real-time GPU fire, smoke, and spark simulation.
Output goes to UE5 via VDB volumes compressed through ZibraVDB, or as flipbook sprite sheets for Niagara particle integration.
Covers the majority of atmospheric and pyrotechnic effects at this production scope.

Real-time GPU fluid and liquid simulation, replacing offline Houdini FLIP for most fluid effects.
Output goes to UE5 via Alembic cache for mesh-based liquid geometry, or as VDB compressed through ZibraVDB for volumetric fluid data.

Offline FX fallback for simulations that go beyond what the real-time tools can handle.
Vellum cloth for complex hero cloth shots as production scales up.
Output as OpenVDB, compressed via ZibraVDB before UE5 import.

Procedural render enhancement, with each AOV layer from UE5 routed through its own dedicated sub-workflow.
Custom Face Scion LoRA trained on FACS turntable renders refines facial detail based on the character's own scan data.
Same operator, same tool, same local hardware as pre-production.

Multi-pass compositing. Assembles AOV EXR sequences from UE5 with ComfyUI-enhanced layers.
Final shot finishing: depth of field, motion blur, grain, chromatic aberration, light integration, and atmosphere.
Output: linear EXR to DaVinci Resolve.

Final colour grade and delivery. Receives the ACES-linearised composite from Nuke intact, with full dynamic range preserved through the grade.
Delivers ProRes and EXR with full ACES metadata.
Review formats converted to display colour space within Resolve.