Phase 1 / May 2026
Forged on Thresholds, built so other VFX teams can use them. 20+ years of supervision, the pipeline I always wanted on a real show, finally shipping. AI closed the gap between knowing exactly how this should work and writing the code, without me pretending to be a pipeline TD.
From script to shot
Shared data contract v2.0.0. Sequence and shot IDs follow the bible. Asset names are PascalCase.
Over the last years I've shifted focus from shot work to workflow optimization and pipeline development; the tools, the workflows, the UX a VFX team actually needs. Claude Code and AI made it possible to build all of it. Before, the gap between “I know exactly how this should work” and “I can ship it as production software” was a five-person engineering team. Now it fits inside one supervisor's head and a weekend of work.
My strength
Workflow analysis, workflow optimisation, pipeline design. Twenty years on the floor, as a VFX & CG Supervisor and former 3D Generalist, spent watching exactly where pipelines hurt, where producers struggle, where clients lose patience. I know what a CG department needs, what a VFX department needs, what's important and what isn't.
What I'm not
I'm not trying to replace pipeline TDs or production engineers. I don't have their coding background. What I have is the supervisor's view of where the workflow actually breaks, and now, thanks to Claude Code, the ability to prototype and ship the fix without waiting for engineering bandwidth that never comes.
The result
Three tools, built by the person who uses them, refined by the producer who depends on them. PXLbid -> PXLflow -> PXLtools, end-to-end, one data contract. Not a product looking for a market, the pipeline I always wanted, now hardened enough to drop into someone else's show.
AI agents trained as a Producer and a VFX Supervisor. They read the script the way I would.
Drop a script in. AI agents, trained as a Producer and a VFX Supervisor, read it, scope it, decompose it into departments, and emit the same v2.0 JSON that PXLflow ingests natively. Re-bidding a revision becomes a delta diff, not a redo.
How it works
is simple as 1-2-3
Counts shots, sequences, assets. Flags complexity. Enough to know whether the project is a 50-shot commercial or a 400-shot feature before you call the client back.
Check, give feedback, change what's needed and proceed to Phase 2, while the tool learns from you.
Continue from Phase 1. The AI re-reads in depth and decomposes every asset and shot into detailed department tasks (model, texture, lookdev, rig, muscle, CFX, etc). Honors your bible, your folder structure, your naming convention, and gives you back a file that you can ingest directly into PXLflow.
A complete department-level breakdown the producer can plan against, not a spreadsheet skeleton you still have to fill in by hand.
Visual interface, company-cost-driven, editable numbers.
The bid generated can be shipped in different formats (XLS or PDF) but the real power is when you ingest it directly into PXLflow as scheduling, not as a static file the producer has to re-type into the planner.
An entire production planned for you, under your supervision.
Client revision lands. Re-run the new script and compare against the old one.
At-a-glance graphs of complexity delta, shot count change, what got added, what got cut and warnings are available for you to review them. Re-bidding now becomes a five-minute conversation.
The tracker you'd build if you'd actually sat in dailies for over twenty years.
Took inspiration from ShotGrid and ftrack for the review side. From Notion and Monday for the planning side. Threw out everything that felt like it was built by someone who'd never sat in a dailies room. What's left is the tracker I always wanted.
I use it daily as the supervisor. My producer uses it daily as the producer. Every feature in there got fought for in a real review, hardened by a real schedule, killed if it didn't earn its place. No one's writing tickets from a roadmap they imagined.
Tasks know what they depend on. Move a deadline, the schedule reflows. Approve a concept, the downstream modeling task wakes up. The kind of behavior that should have been default in production trackers a decade ago.
Per-asset notes, frame-by-frame markup, daily review sessions captured automatically. The dailies room conversation becomes a permanent record without anyone having to write the minutes.
Electron desktop app. Google Sheets and Drive backend. Client-side OAuth. No server to maintain, no SaaS lock-in, no monthly bill that scales with seats. RV integration on Windows. Auto-update via Cloudflare R2.
Independent shippables, sharing a brand, conventions, and a tech philosophy. Each one solves a real production problem on its own. The pipeline emerges when you stack them.
AI-driven VFX script breakdown and bidding. Weeks of script analysis, done in minutes.
VFX production tracker. The best of ShotGrid, ftrack, and Notion-style planning, built by people on the floor. Tracks shots, versions, tasks, dependencies, and approval status — the actual viewing happens in PXLrv.
Production Python toolkit for Maya, Nuke, and Unreal.
Custom made tools for the most used DCCs, developed to solve everyday artist's struggle.
Every tools' development comes from a real production need on the floor. Each one fixes something the DCC should have done out of the box.
Built for Maya 2025 (PySide6) and Nuke 15 (PySide2), versioned releases, install procedures, file headers, and a shared dark UI. Each tile opens to the longer story.
The natural evolution of the turntable scene I released on Gumroad two years ago and that downloaded over 3000 times from artists worldwide.
Now wrapped in a full easy-to-use UI on both Maya and Nuke, end-to-end.
Production-grade quality at a glance.
In Maya: import an asset, load the tool and run an easy-to-follow setup.
You end up with a production-grade turntable in the viewport, with Macbeth chart, reference balls, six perfectly balanced HDRIs (plus 2 free slots for your own HDRI). Render layers come pre-configured with embedded AOVs and shader.
In Nuke: the Maya export drops in cleanly.
The Nuke side handles asset name, side reference, background colour, and output whatever passes you want: colour, beauty, wireframe, all driven by a compact UI and because one render layer (from Maya) contains all the AOVs needed.
Build a client-ready turntable for any assets in a few minutes, not hours/days.
One tool, across two DCCs.
One seamless workflow, hardened on real client work, and the foundation of every other PXLtool that grew around it.
Scan a vendor texture library, build hundreds of PBR materials in your naming convention, in one click.
Every PBR library on the planet uses its own suffix and prefix conventions: _BaseColor, _alb, _diff, basecol, you've seen them all. The PBR Material Creator scans a folder containing hundreds of textures, recognises which ones belong together by shared base name, and generates Maya materials with proper PBR connections in your naming convention.
One click takes you from a chaotic vendor dump to tens or hundreds of properly wired Arnold/Standard Surface materials, ready for assignment. The kind of thing that turns a one-day asset prep into a fifteen-minute job.
GLB import and export with the right twist. Textures extracted, PBR materials built, asset centred, all on one click.
GLB is everywhere now, the standard output for Tripo and other AI model generators, the default delivery for online assets, and Maya doesn't natively handle either side of it well. So I wrote my own.
Import a GLB and the tool extracts every texture, renames them after the asset, files them in the correct sub-folder, builds the PBR material with proper Arnold connections, applies it, and centres the model on the world origin. Export does the inverse cleanly. Built originally to ingest Thresholds-pipeline assets and AI-generated models without manual cleanup every time.
Prototype
The first prototype of Claude integrated directly inside Maya. Drive repetitive tasks by prompt.
An early-stage AI agent integration that lives inside Maya. Ask it to do the boring repetitive thing, renaming, layouting, filing, simple cleanup, and it does it. Preliminary, but already productive on real shots.
The longer-term goal is a Maya-native co-pilot that understands the scene the way a TD does, not just a generic LLM with a viewport.
Beta
Turn a script into a Maya Sequencer build, sequences, shots, frame ranges, ready to previs.
Drop in a script, the tool breaks it down, and builds the entire sequence and per-shot setup directly inside Maya's Sequencer. Frame ranges, shot names, sequence hierarchy, all populated automatically and ready to previs.
Each shot can then be exported separately for downstream work or handed back to a previs animator with the structure already in place. Pairs naturally with PXLbid's script breakdown for the full script-to-Maya pipeline.
Bulk-export a master model split into hundreds of pieces, each as its own OBJ, in one click.
The pain it kills: a master model made of a hundred parts that all need to ship as individual OBJs (3D printing, kitbash libraries, asset hand-off). Maya makes you do that one piece at a time.
This tool does the lot in one click, into a folder you choose, with full export-option control. Includes naming conventions, scale, smoothing, and material handling per piece.
Generate dozens of render layers from selection, auto-named from your group and object hierarchy.
Live-select objects or groups, hit a button, and tens of render layers appear, each named after the object or group it isolates. Built for production scenarios with high-volume layer setups: facial expression studies, lighting variants, hero/secondary/background passes.
Pairs naturally with the Contact Sheet Generator on the Nuke side: render hundreds of layers from Maya, contact-sheet them in Nuke for client review.
Camera matching for Maya, because Maya never shipped one.
A from-scratch camera matching tool for Maya. Match a digital camera to a plate using the standard correspondence workflow, with a UI that's actually pleasant to use compared to the alternatives.
Built because the gap in Maya's native toolkit was costing real production time on every matchmove-adjacent shot.
A complete, robust batch renamer. Simple where it should be simple, deep where production demands it.
Maya's ecosystem is short on full-featured batch renamers, so this one fills the gap. Search-and-replace, prefix/suffix, padding, sequencing, regex, all the operations a real production needs, none of the dead weight.
Designed to be extended as workflow needs come up; new rename operations slot in without rewrites.
Pack dozens to hundreds of renders into a single compact contact sheet, organised by lighting, expression, or asset.
Built for the high-volume look-dev scenarios that no client ever wants to scroll through one frame at a time: thirty-six facial expressions across five lighting conditions, six hero assets in four time-of-day setups, the same character in twenty wardrobe variants.
Render the layers from Maya with the Legacy Render Layer Creator, drop them into Nuke, click once. The contact sheet builds itself, organised by whatever axis you specify, and exports as a single presentation-ready frame. Real client-review economy.
PXLbid emits the v2.0 JSON schema. PXLflow ingests it natively, no glue scripts. PXLtools speaks the same data contract on the DCC side. Sequence and shot IDs follow the bible. Asset names are PascalCase. The same dark editorial UI carries across every surface, web and DCC, because production teams shouldn't have to learn three visual languages to ship one shot.
Shared data contract
v2.0 JSON
PXLbid → PXLflow → PXLtools, end to end. No bespoke per-tool formats. Bible-aligned IDs. PascalCase asset names. The producer's Excel export round-trips back in.
Shared visual language
Dark editorial UI
Same palette, same type rhythm, same component library across all three tools. PySide6 in Maya, PySide2 in Nuke, React 19 on the web, all reading from one design system.
Shared engineering bar
Hand-off ready
Versioned releases, install procedures, file headers on every Python module, backup discipline before every edit. No hacky scripting. Any TD can pick these up on day one.
VFX & CG Supervisor for over twenty years. Former 3D Generalist. Now also a pipeline-optimization consultant: studios that have outgrown their workflow bring me in to look at it the way I'd look at my own.
PXLsuite is the practical answer to a problem most VFX teams run into eventually. The pipeline grows organically, every project bends it a little, and after enough years the workflow no longer fits the way the team actually thinks. The pre-made tools made certain assumptions a decade ago and never revisited them. The team has the experience to fix it but not the engineering bandwidth.
That gap is what changed for me with Claude Code and modern AI. The same supervisor who knows exactly where a tracker should hurt now also has the ability to build the fix in a weekend, harden it on real client work, and ship it as production software. PXLsuite is the proof. Three tools, built by the person who uses them.
If your studio's pipeline used to fit and now doesn't anymore, that's usually a workflow problem dressed up as a tools problem. I look at pipelines the way I look at my own; 20+ years of supervision behind the audit and the ability to build the fix, not just write the slide deck. The door is open.