Skip to main content
Robot arms lift battery modules off a conveyor inside a fenced cell.

Coming Soon

DigiFactory

A factory you can walk through, break, and fix, in a browser tab.

DigiFactory is a digital twin of a fictional EV battery module and pack plant. Every robot, vehicle and worker moves from a deterministic simulation, and every fault comes with a guided recovery you can watch, follow or be tested on.

Explore the plant

At a glance

  • 62

    stations from 29 engineered cell designs

  • 39 / 19 / 94

    vehicles / robot arms / workers simulated in 3D

  • 50

    guided recovery lessons backed by physical faults

  • 20 days

    simulated without a deadlock; one production day in about 8 seconds

  • 7.5 MB

    to load the whole plant in performance mode

  • 160+ fps

    whole plant in performance mode on a gaming laptop GPU

Explore DigiFactory

DigiFactory models a complete 180 m × 120 m battery plant: four module lines feeding a pack line, with storage, logistics and the people who run it. Each station is designed from published engineering sources, then simulated second by second: carriers queue, robots pick and place, stations starve, block and break down. The browser shows it live in 3D, so an engineer, operator or trainee can see what the plant is doing, why, and how to recover when it stops. The plant is fictional; its design rules, rates and procedures are modeled on public sources and labelled as modeled where they are estimates.

  • Module line aisle with carriers, workers and stations.

    Live plant

    The whole hall in 3D, driven by the simulation: carrier robots with the pack at its current stage, articulated robot arms, tugger trains, forklifts and walking workers, at real time or up to 60×.

  • A selected station with live data in the inspector.

    Inspector

    Select any station or machine in 3D, in the plan, in the asset tree or from an alarm, and see its live state, cycle times, counts and data quality, all pointing at the same thing.

  • Assembly journey showing a product's path through the plant.

    Assembly journey

    Follow the product from cells to modules to a finished pack, station by station, with each step’s process and timing.

  • A guided recovery lesson with its target highlighted.

    Guided fault recovery

    When something breaks, the plant shows the physical cause and its knock-on effects. Each fault has a recovery lesson in three modes: Show me, Guide me, and Test me, which refuses unsafe or wrong actions. A lesson only passes when the simulated plant actually produces good parts again. Familiarization content, not certified training.

  • A station fault on the plant overview: the jammed carrier is marked red and the raised alarm tops the alarm list.

    Rewind and what-if

    Scrub back to before a fault and watch it happen again, exactly. Fork the plant, change a decision, and compare the outcome side by side.

  • The guided tour highlights a robot cart and explains it step by step.

    Beginner-friendly guidance

    A Start here welcome, a guided tour of the plant, a Simple layout that hides advanced panels, and an on-screen camera pad, so a first-time visitor can find their way around.

  • An aisle in the high fidelity bay.

    Two visual modes

    High fidelity for photoreal review with baked lighting; a lean performance mode with clean, simple geometry and colors, built to run on a factory-floor laptop and on WebGL2 where WebGPU is missing.

The whole plant

Four module lines feed a pack line, with storage, logistics and people across the hall.

A fly-over across the plant hall, past module lines, carriers and storage.
A fly-over across the plant hall.

Under the hood

How it is built

  • A custom WebGPU renderer written in TypeScript with physically based lighting, baked global illumination and streaming, plus a WebGL2 fallback for performance mode.
  • A deterministic discrete-event simulation with named random streams, so every run and every rewind is reproducible.
  • Plant, cells, robots and the building are generated from data by scripted Blender pipelines. Robot motion uses real kinematics; procedures, rates and dimensions cite their sources or are marked as modeled.
  • Hundreds of automated tests guard the simulation, the data and the renderer.
  • Built with AI-empowered workflows under human direction.

Development

Now

  • Whole plant live in performance mode
  • High-fidelity bay
  • 50 simulated recovery lessons
  • Rewind and what-if
  • Assembly journey

In development

  • The whole plant in high fidelity, with lighting baked per area
  • Conveyors, lifts and doors driven by the simulation
  • Performance testing on integrated-GPU floor laptops

Credits

DigiFactory thanks Rob Tuytel and Poly Haven for the CC0 factory materials. People and motion are derived from the Microsoft Rocketbox Avatar Library under the MIT License. Rendering includes Apache-2.0 work from the Android Open Source Project and Khronos Group, plus LTC data by Eric Heitz, Jonathan Dupuy, Stephen Hill, and David Neubelt.

Third-party notices

Follow the plant as it develops

The public experience is in development.