Maurizio Turoni

CGI Pipeline & Technical Direction

I design production pipelines for CGI teams: the tools, data structures and conventions that let artists find the right file, publish a new version and reuse previous work without losing time. Twenty years in production, from Maya pipelines at Italian animation studios to a complete Blender 5.x pipeline in daily use today.

A Blender pipeline built from scratch

The setting: a small team producing 3D content for several clients at once, with assets reused from one project to the next.

What I built:

• A Python package handling the data layer: projects, clients, assets and their versions, stored in a database.
• A Blender add-on that brings it to the artists: saving, loading and updating props, rigs and scenes directly inside Blender, with automatic versioning
• Naming and folder conventions enforced by the tools rather than left to memory.

Key design choices

One source of truth. A single MariaDB database on the studio NAS, reachable over VPN for remote work. The database has the final word on versions, statuses and relationships: everything else, file names on disk included, must conform to it.

Saving is not publishing. Every save from Blender creates a numbered version, and numbers are never reused. Versions move from WIP to Review, Approved and Published only through the web dashboard, where every access requires a login and every status change is kept in a history that cannot be edited.

Every scene is a snapshot. On save, a scene records the exact version of each asset it contains. Its history stays true as assets move on, and it is possible to know in advance which scenes an update will affect.

Checks where constraints can’t reach. The integrity tools verify the references the database alone cannot protect, and were tested on a copy with deliberately introduced faults.

A real case. A model and its rig, sharing the same name, were being recorded as a single asset: for months scenes worked through the wrong record, and nobody noticed. The fix separated the data, added a constraint and introduced the checks that would now catch it automatically.

Earlier work

Pipeline development and asset management systems at Maga Animation and Rainbow CGI, production tools written with the Maya API and PySide2, and stereoscopic supervision on CGI projects. Some technical notes from those years:

• Optics budget in a Maya scene
• Stereo 9 views camera rig
• Qt5: C++ versus Python

AI, used with judgement

I use AI development tools daily, keeping architecture and review in my own hands. On a personal basis I am working on a LoRA for the Wan 2.2 model, aimed at reducing artefacts in AI-generated stereo content.

Contact

maligut64@gmail.com · linkedin.com/in/maurizioturoni