Three parts: what has actually changed, ComfyUI from zero, then the commercial.
Maya for the control, ComfyUI for the speed. Nothing else to buy.
No ComfyUI experience assumed, and no model training or theory.
CG holds what must be controlled, AI supplies what is expensive.
Where AI actually fits in a CG pipeline, and how to keep control of the shot when you use it.
Maya holds the camera, the framing and the product. AI supplies surface, light and speed.
The toolset
The curriculum
Three parts: getting your bearings on what actually changed, ComfyUI taught properly from zero, then the commercial built end to end.
The landscape without the noise: why there are so many models, why they keep changing, and why most of that churn does not matter to a working artist. What has genuinely changed in production, and what has not.
Enroll to unlockWhere AI output is accepted today and where it is refused, and why "it depends on the client" is the truthful answer rather than a dodge. The phases where it is already a clear win, and the phases where CG still has to do the work.
Enroll to unlockCG holds what must be controlled, AI supplies what is expensive. Camera, geometry, product accuracy and continuity stay in CG. Surface realism, light exploration and iteration count go to AI. The split is a production decision, not a technical one.
Enroll to unlockInstall and first run, from zero. What the pieces are, getting a model in place, and generating something. Deliberately no theory about how the models work, because you do not need it to use them.
Enroll to unlockNodes, inputs and outputs, and why a graph beats a text box: every stage becomes something you can see, change and re-run. Including how to read a graph somebody else built.
Enroll to unlockOrganising a graph so it survives next week: grouping, naming, the inputs you expose deliberately, and setting a shot up so you can run ten variations without rebuilding anything. A procedural approach, applied to AI.
Enroll to unlockStaging the product and the camera in Maya for a commercial frame. Why a grey, unlit, untextured playblast is the ideal input rather than a compromise: it is the one thing a prompt can never give you.
Enroll to unlockThe two-source discipline. The playblast is spatial only, and its shading and lighting must be actively ignored. Real product photos are the only authority for material, roughness, weave, stitching and edge wear. How to check the lock held instead of assuming it did.
Enroll to unlockPass 1 output becomes the locked master plate: camera, object, materials and colour must not move. Style references influence lighting, background and depth of field only. Relighting the same shot repeatedly without it drifting, which is the whole reason for splitting the passes.
Enroll to unlockDriving the locked frames into movement, and holding consistency across shots. Taught model agnostic and demonstrated on more than one engine, so the method outlives any single tool.
Enroll to unlockAssembling the shots into something that reads as an ad rather than a reel of tests. Pacing, sound, and the finish.
Enroll to unlockBack to module 02 with the work in hand: what you would now show a client at concept stage, what you would still build in CG, and how to slot this into an existing pipeline without betting a delivery on it.
Enroll to unlockIn production
The full module list below is locked. Recording is in progress, and everyone in the Discord hears the release date first.
Project files included · Private Discord · All future updates included, free, forever
Software used: Maya and ComfyUI, plus whatever you already cut video in
Skill level: Beginner
Course type: Pre-recorded, self-paced
Status: In production, recording now
Instructor: Cristian Spagnuolo, VFX & CG Supervisor, Mentor & Beta-Tester
Maya, or another package you can stage a product and a camera in
A GPU that can run ComfyUI locally, or a cloud instance if yours cannot
Enough 3D to frame a shot and publish a playblast
No ComfyUI experience. The setup is taught from zero
No model training and no maths. Off the shelf models only
This course is made for:
CG artists who want to know where AI actually belongs in a pipeline, and where it does not
VFX and advertising professionals who have to justify the approach to a supervisor or a producer
Freelancers and students who need to answer a client at concept stage in hours rather than days
Anyone who has never opened ComfyUI and would rather learn it properly than get a lucky result
The boundary is part of the course rather than something it talks around. AI collapses the early phases every time: concepting, storyboarding, lighting concepts, look dev exploration. Whether it can deliver a final commercial depends on the client, and some are exacting enough that the answer is no. Being straight about that is what makes the approach usable in front of a producer.
Google