AI Training · Synthetic Data · VFX Pipeline Construction · 20 Years in Production
Lockehaven AI is a production construction studio that builds, integrates, and delivers AI training pipelines, synthetic data systems, and VFX infrastructure — from pipeline construction to neural rendering to end-to-end digital human localisation.
What We Do
We embed as a production partner — scoping, building, and delivering. Whether you need a Nuke pipeline constructed from scratch, an AI training system stood up, synthetic data built for your ML team, or a complete VUBbing workflow for a streaming release, we do the actual work.
End-to-end pipeline architecture and delivery. Nuke 17.x compositing systems, multi-layer EXR workflows, ACES/OCIO color pipelines, and show-agnostic config systems that scale from episodic TV to feature film.
See Capabilities →Production-ready AI training pipelines across DeepTracker facial systems, synthetic data construction, 3D Gaussian splatting, NeRF reconstruction, and PyTorch / ComfyUI tooling — built for studio-scale throughput.
See Capabilities →Complete VUBbing pipeline delivery — from facial tracking and 53-blendshape systems through Spede neural rendering, wire removal, and final compositing. Proven on major streaming titles.
See Capabilities →Unreal Engine pipeline engineering, Unity VR (Meta Quest 3), in-camera VFX, USD/Karma/Houdini workflows, and Gaussian splatting for interactive and on-set productions.
See Capabilities →Engagement Model
We operate as a senior production partner embedded in your team. No junior handoffs. No account managers between you and the work. Every engagement runs through the same experienced technical leads who scoped the project.
We audit your current pipeline, review deliverables, and define exactly what will be built. Fixed scope or iterative — your call.
System design docs, tool selection, and integration mapping before a single line of code is written. We price per milestone, not per hour.
We build in your environment — or ours. Python, C++, VEX, C#, AWS. Production tools that work on day one without babysitting.
Round-trip validation, error-threshold testing, and cross-show stress testing. We don't hand off until the numbers are right.
Full documentation, team training, and optional ongoing support retainer. Your crew should own what we build.
Technical Depth
Twenty years of production-proven tools and workflows. We only build with software we've shipped with — every capability listed below has been delivered on broadcast, streaming, or feature productions.
Nuke 17.x pipeline development, gizmo & plugin authoring, DeepTracker integration, multi-layer EXR workflows, ACES / OCIO color management, Syntheyes camera tracking, stereo compositing.
Large-scale environment pipeline, SpeedTree ecosystems, Houdini procedurals, lighting lead / CG lead supervision, bidding & pitching support, 20 years episodic TV & film delivery.
Unreal Engine pipeline engineering & previs, Unity VR (Meta Quest 3), Gaussian splatting (GSOPS / COLMAP), USD / Karma / Houdini integration, in-camera VFX & on-set supervision.
Redshift, Arnold, VRay, Renderman, Mantra, Cycles. Look Dev / Lighting TD services. Substance-based material pipelines. Marvelous Designer cloth simulation for digital humans.
Vertex-based rig conversion (Python), blendshape systems (53-shape additive merge), Houdini FX simulations, Maya rigging, procedural asset creation & validation pipelines.
Python (Nuke / Houdini), C++, VEX, C#, AWS S3 / EC2 integration, Linux (Rocky Linux), Git / GitHub, show-agnostic YAML config systems, batch processing & automation at scale.
AI Training · Synthetic Data
We build production-grade AI training pipelines and synthetic data systems — not proof-of-concepts. Our work ships to broadcast and streaming platforms with the same reliability standards as any other VFX deliverable.
DeepTracker system design and deployment, face tracking refinement in Nuke, neural rendering pipeline construction, 3DMM / blendshape model integration, expression coefficient export, identity EXR generation.
Synthetic data pipeline construction, Houdini data visualization for ML scientists, SAM2 mask propagation, segmentation-to-Nuke Bezier curve automation, ViTMatte alpha extraction at scale.
NeRF pipeline build and integration, 3D Gaussian splatting workflows, COLMAP photogrammetry, PLY sequence processing, PyTorch3D coordinate system alignment and validation.
PyTorch / PyTorch3D, ComfyUI / LoRA fine-tuning, Veo & Nano Banana video generation pipelines, Claude Code agentic workflow development, Spede / centipede DAG orchestration.
Digital Human Localisation
We deliver the full VUBbing stack — facial tracking through neural rendering through final comp. Our pipeline has been built and refined in active production on major streaming titles. We can stand up, extend, or take over an existing workflow.
DeepTracker build (Nuke plugin architect), expression coefficient pipelines, identity offset EXR output, 53-blendshape additive merge networks, baked keyframe export, round-trip validation to error < 1e-6.
Spede NR resubmit pipeline, FlexibleExport / Import gizmo development, camera pose back-conversion, S3 dataset packaging & upload automation, Spede DAG orchestration and monitoring.
SAM2 → RotoPaint automation pipeline, DMV inpainting via Veo, GPT-4o / Moondream wire click detection, SAM2 track+segment workflows, ACEScg / sRGB compositing and delivery.
Shot config / YAML system architecture, localize + build shot trees, multi-show Nuke template systems, batch EXR / NPY processing, AWS S3 upload automation and job monitoring.
Track Record
Twenty years of delivered work across broadcast, streaming, and feature film — from flagship genre television to global theatrical releases to AI-driven digital human productions.
The People Behind the Work
A combined century of knowledge and experience. The best and brightest of the Film, Television and Commercial industry.
20 years across episodic VFX, feature film, and AI-driven digital human systems. Architect of DeepTracker, Spede neural rendering pipelines, and Gaussian splatting workflows in active production.
Seasoned VFX supervisor with deep episodic and feature experience. Specialises in on-set supervision, sequence oversight, and bridging the gap between shoot and final deliverable.
Senior compositor with extensive Nuke pipeline experience across broadcast and streaming. Brings precision and speed to complex multi-layer EXR and ACES colour-managed workflows.
Technical Director with a focus on pipeline architecture and automation. Builds the tooling and infrastructure that keeps productions running at scale — on time and without surprises.
Environment and look development specialist with a strong eye for large-scale CG worlds. Experienced across Houdini procedurals, SpeedTree, and lighting pipelines for episodic television.