About LOFT
The OS for AI Production
LOFT is built by LOFT, founded in 2022 in Berkeley, California by longtime business partners Gary Mundell, CEO, and Remko Noteboom, CTO. We build production infrastructure for the AI era.
We built this once already
We were members of the original Alias|Wavefront Maya development team. Noteboom then built Nelnet at Nelvana - the world’s first digital production workflow tool - and went on to develop TACTIC, deployed across 350+ studios. That work earned a 2021 Technology & Engineering Emmy Award.
LOFT itself began as a resource and production management tool built with Jellyfish Pictures, deployed studio-wide to hundreds of users, where it transformed their business model with measurable efficiencies across production and finance. That system is the direct predecessor to LOFT as it exists today.
Across 25 years we have lived through every major shift in production: analog to digital, local to cloud, software to agentic orchestration. One problem has proven the most persistent - keeping film and media assets organized. Now we are doing it again, for AI production.
The gap
Generative models are becoming commodities. The systems that turn their outputs into production assets - traceable, versioned, accountable, and connected to human intent - are not.
Studios can generate imagery today. Many still cannot say which prompt and model produced a frame, who approved it, or how to reproduce it. Provenance has stopped being optional: streamers, buyers, and legal teams increasingly expect an account of what contributed to a delivered frame.
LOFT: the operating system for AI production
LOFT is an Integrated Production Environment - not a model or a generation engine, but production infrastructure. It connects to the models a studio chooses and provides the durable structure beneath them.
Just as an IDE doesn’t ship your compiler, LOFT doesn’t ship your models. It hosts the models an organization chooses while providing the layer every service plugs into: assets, lineage, provenance, and a shared intent-to-action record. Models are interchangeable. The production record is not.
In stop-motion, the armature is the steel skeleton inside the puppet. The audience never sees it, and nothing stands up without it. LOFT is the armature of an AI production. The models are the surface, and the surface keeps changing.
Serious infrastructure, without the learning curve
LOFT is professional-grade capability with a floor low enough that anyone can use it.
It scales to feature films, episodic series, animation, VFX, games, and enterprise creative operations - projects measured in hundreds of shots, teams of dozens or hundreds, months of review cycles and sign-offs. At that scale, the questions that matter are ones ordinary tools don’t answer: which prompt produced the approved frame, who signed off on version four, what the sequence cost, how this take connects to the cut, and whether any of it can be reproduced six months later.
But depth is not the same as difficulty. Most production software asks you to learn its model of the world before you can do anything with it. LOFT works the other way: describe what you want, provide a reference, and it coordinates the models and steps behind you. The complexity is real - it just isn’t yours to manage.
Where we’re different
Intent drives the pipeline. Creators describe what they want and provide references through a conversational assistant. LOFT coordinates the models and production steps, recording every prompt, decision, and iteration as it goes.
One system, not a stack of tools. Asset management, orchestration, generation control, review and approval, cost tracking, editorial, governance, and delivery in one place. Editorial is a first-class asset that references its source material, so a new take re-points the cut instead of forcing a manual reconform - provenance survives into the edit, the place traditional pipelines drop it.
Neutral by design. Models are interchangeable services around a stable production core. As providers, capabilities, and pricing churn, the production record remains intact.
Resource planning built in. Budgets, timelines, scheduling, and resource management are native options, so the same system that tracks creative work answers what it costs and when it lands.
The seam is the problem. Most studios bolt AI onto a traditional pipeline stage by stage. The gains are real, but most of production’s cost lives between departments: communication, versioning, review cycles, format conversion, the constant translation of intent across teams and tools. Speed up one stage by 10x and the shot still takes nearly as long, because the bottleneck was never that stage.
Forged in production, not on a whiteboard
LOFT was developed 100% outside of Tippett Studio, in independent collaboration with Phil Tippett, on productions that each exposed the next problem:
Sentinel - Tippett’s next stop-motion project, unveiled at Cannes’ Frontières in 2024 as a teaser for a film then in development. Generation was the easy part; one idea became dozens of variations across models and people, and tracking what we learned was hard. That finding is the origin of the product.
Pequin - a finished one-minute piece, cut and versioned entirely in LOFT, run through editorial deliberately to design out Sentinel’s wounds: version chaos and upload/download churn.
Mad God 2 (Mad Goddess) - the sequel to Phil Tippett’s cult feature Mad God, with a teaser managed end-to-end in LOFT. The production is currently seeking participants and funding.
Phil Tippett - two Academy Awards, the creature and animation work behind Star Wars, RoboCop, Jurassic Park, and Starship Troopers - serves as advisor to LOFT. He is not a figurehead: the requirements that became LOFT were written alongside his films.
LOFT and Tippett Studio are unaffiliated. LOFT is a LOFT product, conceived and developed independently, and available to any studio. We continue to work with Phil on independent productions, which is what keeps the product honest.
What we believe
The moat is the record, not the model count. Aggregating models is easy. Building the system of record that turns their outputs into traceable production assets is not.
Provenance is not a feature. It is the foundation of accountable creative work, and it belongs in the pipeline from the first prompt - not reconstructed after delivery.
Human intent stays at the center. AI can accelerate individual tasks. It cannot carry authorship. Our job is to make sure it never has to.