Contact Us
GROUND TRUTH IS READ, NOT ESTIMATED

Every video frame ships with the exact world state the simulation computed at that instant — every entity, every gameplay component, every world resource. Nothing on this page is inferred from pixels; the numbers come straight out of the engine.

ONE CLOCK, EVERY MODALITY

Video, player input, camera pose, full world state, the event log, entity-ID masks, metric depth and 2D boxes all share the 60 Hz simulation tick. Frame N and every data row with tick: N describe the same instant — aligned by construction, not synced afterward.

CAUSAL EVENTS, NOT CORRELATIONS

Every game rule that fires is logged with entity ids and parameters: input fire → rule Shot → rule Damage (−24 HP) → rule Death. Per-moment causal annotation of who did what to whom — supervision that reconstruction-based datasets cannot supply.

DETERMINISTIC & FORKABLE

Fixed timestep, seeded, per-tick digests: any episode replays bit-identically from its action stream, and any tick can be forked into counterfactual branches with exact ground-truth futures — the answer key action-conditioned world models can't get from real footage.

ENGINE-NATIVE GAMEPLAY DATA

A game recording that carries its own answer key — per-tick world state, a causal event log, per-pixel labels, and the player’s every input — all read from a deterministic simulation on one clock, never estimated from pixels.

Sample episode ep001 · Urban Block — 32.4 s · 60 ticks/s · 1,946 frames · 47 entities · combat, pickups, NPCs, wanted system. Every stream on this page is the real file.
Overlays All overlays are the shipped label files, redrawn per tick — with SEG on, click any object to decode its pixel → entity id → full state
RENDERED GAMEPLAY
TICK 0000 · digest
tick 0 / 1945
Input @ tick no keys AIM FIRE idle look Δ 0.00 / 0.00
AlignmentVideo frame N = simulation tick N = every JSONL row with tick: N. No drift, no interpolation — alignment is a property of the export pipeline.
Schematidel-sim-episode/0.2 · deterministic serialization · per-tick digests
SourceProprietary ECS-native engine. Adaptive scripted driver acting through the real input surface; human sessions record through the same interface.
Cross-modality identity
A pixel resolves to the full state of the thing it depicts

Entity ids are shared across the state stream, the event log, the segmentation masks and the bounding boxes. There is no detection model and no labeling step between a pixel and the truth about the object it shows.

1Toggle SEG above and click any object in the frame.
2The mask pixel decodes to an entity id — id = R + G≪8 + B≪16 + A≪24, straight from the shipped PNG.
3That id keys the same tick's state row: the full component JSON opens in the panel, and the same id appears in every event that touches the entity.
Deterministic replay · counterfactual branching
Fork any tick, get an exact alternative future

The simulation is a deterministic function of state and input: fixed timestep, seeded, digest-verified per tick. Load the state at any tick, feed a different action stream, and the engine produces the precise ground-truth future of that branch — states, events, and re-rendered video.

recorded episode · actions.jsonl tick 410 fire = true → Shot → Damage −24 → Death @429 fire = false → target alive, heat stays Calm same prefix · exact divergence

Two branches share an identical history and differ by one input flag — the divergence of their futures is pure causation, with an exact answer key. This is the benchmark real footage cannot provide: same past, different action, both futures known.

Production delivery
Everything on this page is customizable per engagement
TierContent
Gameplay episodesVideo (MP4/PNG), per-tick actions, camera, full world state, event log. Batch-scale multi-episode runs with consistent schema and per-episode reproducibility metadata.
Per-pixel labelsEntity-ID segmentation and metric depth available today; semantic-class segmentation, surface normals and optical flow scoped per engagement.
3D assetsScenes, characters, vehicles and props as inspectable glTF/GLB, cross-referenced to the entity ids that use them.
Custom scenariosObjectives, agents, hazards, mechanics and environments authored to a specific research question; export schema defined to specification.
Simulation accessCounterfactual rollouts and branch-from-any-tick evaluation setups, available on request.