Datasets:
Request sample access
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This repository contains the dataset specification, annotation schema, and sample metadata files. The production dataset (10,000 hours of rights-cleared gameplay trajectories with synchronized video, inputs, and game state) is delivered directly under a commercial license. Approved requesters get the full trajectory schema and data dictionary in this repository, and can request a review package with real trajectory samples and QA summaries. Requests are reviewed within 1 business day.
Log in or Sign Up to review the conditions and access this dataset content.
Action-Conditioned Gameplay Dataset
10,000 hours of rights-cleared gameplay trajectories with synchronized video, keyboard/mouse/controller inputs, camera motion, player state, object state, events, goals, rewards, and outcomes for world models and AI agents.
This repository contains the full technical specification, annotation schema, and sample metadata files (Parquet). The production dataset is rights-cleared and delivered directly to buyers. Request access to see the full schema and get real trajectory samples.
Overview
The Action-Conditioned Gameplay Dataset is a 10,000-hour collection of rights-cleared gameplay trajectories designed for world model training, AI agent learning, action-conditioned video prediction, interactive environment generation, and state-action-outcome modeling.
Unlike passive gameplay videos, this dataset captures synchronized gameplay footage, player inputs, camera movement, game state, event logs, task goals, and outcomes. Trajectories are structured so models can learn the relationship between the current visual state, the chosen action, and the resulting future state, supporting long-horizon prediction, inverse dynamics, policy learning, game AI, interactive simulation, and environment modeling.
At a glance
| Total volume | 10,000 hours of trajectories |
| Episodes | ~500,000 gameplay episodes |
| Video frames | 1B+ rendered frames at source frame rate |
| Input events | 300M+ timestamped keyboard, mouse, and controller actions |
| State snapshots | 50M+ structured screen-state or game-state snapshots |
| Rights status | Rights-cleared for commercial training |
Environment coverage
First-person navigation, third-person action, platforming, driving, puzzle solving, resource collection, object interaction, inventory management, exploration, and multi-step goal completion.
Technical specifications
- Per-trajectory data: rendered gameplay video, keyboard/mouse or controller inputs, action timestamps, camera pose, view direction, player position, velocity, inventory state, health or status indicators, object interactions, environmental events, goal progress, reward signals, success/failure labels
- Subset-dependent extras: depth maps, semantic segmentation, object bounding boxes, minimap state, NPC positions, collision events, physics state, per-frame captions
- Annotation coverage: action timestamps, camera pose, view direction, player position, velocity, inventory, health/status, goal progress, reward signals, event labels, success/failure
- Buyer evaluation metrics: action-state synchronization, frame rate, input frequency, state frequency, episode length distribution, goal completion labels, reward quality
- QA metrics: rights-clearance review, corrupted episode filtering, sync validation, duplicate episode removal, metadata consistency checks
- Delivery format: gameplay video, input logs, state logs, event logs, CSV metadata, JSON trajectory files
Trajectory schema
The gated file annotation_schema.json in this repository contains the full trajectory schema with an illustrative example record, including the per-timestep state-action structure and reward/goal signals.
The samples/ folder holds sample metadata in Parquet format: index_sample.parquet (item-level) and states_sample.parquet (event-level), both conforming to this schema. Values are generated to illustrate structure and field distributions; production records ship in buyer review packages.
How to evaluate this dataset
- Request access using the form above. Requests are reviewed within 1 business day.
- On approval you get the gated files in this repository: full trajectory schema, sample metadata Parquet files, data dictionary, and access instructions.
- Request a review package and we deliver real trajectory samples (video, input logs, state logs), QA summaries, and licensing documentation within 2 business days.
All samples are delivered with structured CSV metadata and JSON annotation files where available. Buyer review packages include representative media files, metadata samples, annotation schema, QA summaries, and data dictionary documentation.
Licensing
The production dataset is rights-cleared for commercial AI training and licensed directly by Datoric with rights-clearance review and chain-of-custody documentation. Subset (by environment type or hours), exclusive, and custom-collection options are available.
About Datoric
Datoric supplies rights-cleared, spec-exact training data for frontier AI labs and enterprise model teams: gameplay trajectories, computer-use traces, egocentric and industrial video, human manipulation data, robot episodes, and expressive multilingual voice. We also run managed collection pipelines for custom specifications.
Contact: nikhil@arzule.com
- Downloads last month
- -