I build the tools I wish I had. Every HDA, script, and pipeline system here was born from real production needs: problems I hit on the floor that I turned into reusable solutions.
Full PySide6 desktop application built for the studio. A single interface handling denoising and render pass combining, AYON publishing, Deadline farm job submission and monitoring, MP4 generation with burn-ins, a collaborative infinite canvas with network sync, and shot cleanup tools. Features a tab-based architecture with worker threading, event bus for cross-tab communication, thread-safe state management, and network-centralised logging. Deployed across the studio's 50+ person team.
The built-in ComfyUI tab connects to a farm-hosted ComfyUI server via Deadline, giving every artist access to generative AI workflows without local GPU requirements. Supports image editing and generation (Qwen Image Edit), image-to-video (LTX Video), single and multi-view image-to-3D mesh generation with .glb export (Trellis 2), text-to-motion capture with .fbx output (HunyuanMotion), normal map extraction (Normal Crafter), and AI audio generation. Includes admin-managed model and workflow categories, per-workflow analytics with node-level timing, a shared gallery with network sync, and a prompt builder for artist-friendly text inputs.
Custom AYON server addon for automated render denoising and pass combination. Integrates with the studio's AYON pipeline as a server-side addon with both client and server components, handling denoising workflows at the pipeline level rather than per-artist.
Designed and built the full integration layer between ComfyUI and the studio's production pipeline. Artists access generative workflows through a dedicated tab in Luma Tools, with all inference running on the render farm via Deadline. The system supports hot-swappable workflow categories managed by admins, per-node analytics for performance profiling, and a shared gallery with favourites synced over the network.
Current production workflows include concept-to-render image editing, image-to-video generation, single and multi-view image-to-3D mesh export (.glb), text-to-motion capture export (.fbx), sequence-based normal map extraction, and AI audio generation. New models and workflows are continuously evaluated and integrated as the research landscape evolves.
Active monitoring of AI research papers and open-source implementations relevant to VFX production. Tracking 30+ papers across categories including image editing, 3D generation, motion synthesis, depth estimation, and video generation. When ComfyUI node implementations become available, workflows are tested and integrated into the studio pipeline within days.
Full USD pipeline architecture built on AYON (formerly OpenPype). Custom Asset Resolver implementation with version pinning, Houdini LOPs/Solaris workflows, and farm-aware publishing. Handles the full lifecycle from asset creation through render submission to final delivery. Production-proven across Shaka iLembe (600+ VFX shots), Headspace, and Kings of Joburg S3.
Active contributor to AYON's open-source ecosystem. Contributions include adding support for locally pre-installed USD resolver binaries with farm rendering support (ayon-usd #106), fixing Windows drive letter parsing for project roots in the C++ dev tools (ayon-cpp-dev-tools #10), and adding long path guards for Windows clients in the bridge client (ayon-bin-bridge-client #22).
Custom Deadline integration for automated render submission, pass building, and publish hooks. Environment management, JSON metadata generation, and resolution detection for seamless farm rendering at scale. Supports Karma, Blender Cycles, and Maya renderers across the studio's 50+ person team.
Collection of Python scripts and tools for studio workflow automation, including VEX snippet management, network creation utilities, viewport camera controls, file format conversion, and custom shelf tools. Built to remove friction from daily artist workflows.
Interested in the technical details?