Artificial World Intelligence™
    What we are doing to build Artificial World Intelligence™

    Training data,
    captured from worlds.

    The world's richest multimodal data set, fully licensed. 25 modalities across three tiers, captured to your spec.

    For AI labs

    The richest multimodal training data anywhere - engine telemetry, HUD-less video, and more.

    See our data
    For IP holders

    Video game publishers, developers, and content studios - turn your IP into recurring revenue.

    Monetize your content
    Partners
    25+
    Exclusive game publisher partnerships - and growing daily.
    Titles
    80+
    Licensed titles across genres, with new captures coming online weekly.
    Modalities
    25
    9 Core. 6 Add-ons. 10 Premium.
    Catalog
    100K+
    Hours of licensed capture.
    Data & assets
    1 PB+
    More than 1 PB of data and assets.
    Trust
    100%
    Licensed at the source. Audit-ready provenance, per-asset rights trail, opt-outs.

    02/Live capture
    Explore the signals,
    alongside every frame.

    See our capture capabilities on real licensed gameplay. Explore RGB, depth, camera and input signals in the viewer, then scope your capture across all 25 modalities.

    03/Library + spec
    Explore what we have.
    Capture what you need.

    25+ publisher partners. 80+ licensed titles across a wide range of game genres, plus 3D environments. Growing daily. Don't see your target genre, mechanic, environment, or modality? We capture to spec.

    25+Publisher partners
    80+Licensed titles
    100%Captured for AI training specs

    Games

    Engine-level capture from AAA and indie titles. First-party exports with action labels, camera telemetry, depth, normals, and game state.

    By size
    AAAAAIndie
    By genre
    FPSOpen-worldSurvivalRPGActionRacingCo-opSimulation

    3D Environments

    Worldbuilder libraries with full scene graphs. USD, FBX, glTF assets ready for simulation, training, and synthetic data.

    By type
    UrbanNaturalIndustrialInteriorFantasySci-fi

    Need something we don't have yet?

    Share a spec, an RFP, or just your training goal. We design capture plans around your requirements - across any title in our catalog, or one we license for you.

    Request a custom capture
    04/Why it matters

    Scraped video is a dead end. Engine data is the breakthrough.

    Pixels are the surface of a world. Explore 25 modalities across Core, Add-ons and Premium, with capture and Services scoped to your spec.

    a.
    Per-pixel ground truth

    Metric depth and Surface normals are Add-ons. Segmentation is Premium. Scoped per order, not inferred from pixels.

    b.
    Action + camera traces

    Every input, every camera pose, every game-state tick - aligned to the frame.

    c.
    HUD-free rendering

    Rendered with UI disabled. Clean signal for world-model training.

    d.
    Provenance chain

    Every frame traces back to its licensed source and rights holder.

    Capture capabilities

    25 modalities / 3 tiers

    CoreIncluded in every hour

    9 core modalities in every hour, 60 fps, frame-synced on one clock, 100,000+ hours ready to ship.

    • Pre-HUD video
    • Post-HUD video
    • Audio
    • Keyboard inputs
    • Mouse inputs
    • Camera telemetry
    • World state and physics
    • Player position and pose
    • VR inputs (VR titles)

    Add-onsOptional · Scoped per order

    6 modalities, chosen a la carte, ready on part of the catalog and captured to order for the rest.

    • In-engine action events
    • Skeleton
    • Optical flow
    • Metric depth
    • Surface normals
    • Collision

    PremiumOptional · Scoped per order

    10 modalities, scoped and captured per order.

    • Segmentation
    • Motion vectors
    • Multi-camera rigs
    • Multiplayer sync
    • Albedo
    • Multi-layer audio
    • Free camera
    • Task instructions
    • Cube maps
    • Material segmentation

    ServicesScoped separately

    Scoped per order. Player webcam and mic require written player consent.

    • Custom tasks and scenarios
    • Statistical coverage analysis
    • Player webcam and mic (with written player consent)

    Core is priced per player-hour. Add-ons, premium modalities and custom capture are quoted per order. Request pricing and we'll scope it with you.

    Inspect sample metadata · scene_00472.json
    scene_00472.json
    LICENSED
    {
  "scene_id": "ol_00472",
  "source": {
    "studio": "[partner]",
    "title": "[cleared IP]",
    "license_id": "OL-2026-4412",
    "rights": "train + publish"
  },
  "capture": {
    "rate_hz": 60,
    "duration_s": 842.3,
    "resolution": "1920x1080",
    "hud": "disabled"
  },
  "core_modalities": [
    "Pre-HUD video",
    "Post-HUD video",
    "Audio",
    "Keyboard inputs",
    "Mouse inputs",
    "Camera telemetry",
    "World state and physics",
    "Player position and pose",
    "VR inputs (VR titles)"
],
  "add_ons": ["metric_depth", "surface_normals"],
  "metadata": {
    "locale": "urban_night",
    "weather": "rain",
    "narrative_tags": ["chase", "first_person"],
    "actor_count": 3,
    "physics_sim": "full"
  },
  "provenance": "sha256:9f2c..."
}

    05/Engine vs. web
    The visible scene.
    The world behind it.

    A video shows what happened. Capture from the engine can also record the geometry, inputs and world state behind it. Direct signals, not estimates made from pixels.

    Inside a dataset / Game Recordings v44 visual signals shown
    Game Recordings v4 multi-title montage. Each of four scene groups has clean RGB at top left, player-view RGB at top right, surface normals at bottom left and depth at bottom right.
    Actual dataset montage, across multiple titles. Select a signal to highlight its panels. Static preview, not live telemetry.
    Inspect the visual signals
    Add-ons / Optional

    Metric depth

    Per-pixel depth from the renderer, rather than depth estimated after capture. The live viewer above includes an RGB and depth comparison.

    What this enables
    Connect the rendered image to scene distance for spatial understanding and reconstruction.
    From web video

    Observe the outcome.

    Visible imagery and sound, when present. Engine depth, camera poses and player inputs are not recoverable as original engine records from pixels alone. Training rights need separate verification.

    From licensed engine capture

    Train on the underlying signals.

    Pair the rendered world with frame-synced engine context and recorded inputs. Scope optional modalities to your research, with a license from the rights holder.

    25 modalities across three tiers. Capture scoped to your spec.

    • 09CoreIncluded in every hour
    • 06Add-onsOptional, a la carte
    • 10PremiumOptional, per order
    • ServicesSeparate, scoped per order
    Explore a real capture

    Build AI on real worlds.

    Whether you're a frontier lab training the next generation of models, or an IP holder ready to unlock your catalog - Origin Lab is the layer between you.

    I'm an AI labI'm an IP holder