Skip to content

OWL Control is a desktop application that records gameplay footage and input data from video games to create open-source datasets for AI research.

License

Notifications You must be signed in to change notification settings

Wayfarer-Labs/owl-control

Repository files navigation

An Owl emerging from an abstract globe of the Earth, all floating above a hand

OWL Control

Help train the next generation of AI by sharing your gameplay!

License

OWL Control is a high-performance Windows app that captures control data from games. These datasets are fundamental to training world models that power sophisticated robots and simulations.

About

We carefully log keyboard, mouse and gamepad inputs from the active game to a file synced with a mini video of the game. No other windows or control input is recorded. Any other window or input - including any microphone or camera - is not captured.

OWL Control is fully open-source, so anyone can verify its inner workings by reading the code or feeding this page's link to your favorite AI language model. The software is developed and enriched by a vibrant community on Discord, and anyone is allowed to contribute to the project

System Requirements

  • Windows device capable of running games at 60 FPS.
  • Keyboard, mouse, trackball, trackpad, Wired/Wireless XBOX or Wired PS5 gamepads. PS4 controllers may be used with DS4Windows.
  • A reliable internet connection. Uploading may take a long time.
  • Computer games! We provide eligible games to veteran players!

Setup

Watch the Walkthrough Video or follow the steps below:

  • Download OWL Control installer.
  • Run the installer.
  • Launch the app from your desktop or Start menu.
    • Check the bottom right corner of your screen for the turquoise OWL control icon. The app may already be open.
  • Create an account at our website. The link is also in the app.
  • Generate an API key.
  • Copy your API key into the app and click Continue.
  • Review the terms of recording. Only record if you agree with them.

Important

  • We don't capture your microphone or anything outside the active game.
  • We screen and filter all the data we receive, and any private information is removed.
  • We will freely release our scrubbed and prefiltered data to the research community under permissive and open-source license.

Usage

  • We accept recordings of these games in PvE modes only.
  • Hit F5 key to switch recording on and off. Please only trigger this within a game you want to record.
    • A small overlay on a screen corner shows that the app is open and recording.
    • Position, keys, and goose-flavored notifications can be customized.
    • If your game runs slowly while recording, lower settings in Video Encoder, or lower the game detail or resolution.
  • Recordings will be tracked in the app. Recordings ready to be uploaded are marked in yellow.
  • Recordings may be too short or not have enough activity to submit. These recordings are marked with red and tagged invalid.
    • A message why they can't be accepted will appear. This information can be useful to share in our Discord server.
  • You can review your recordings by clicking its number. A window will open showing the folder contents.
    • Non-video files in this folder can be opened in Notepad or other text editors.
    • Location of the entire recordings folder can be changed with the Move button to the right of Upload Manager
  • Upload recordings by hitting the Upload Recordings button.
    • If your connection is slow, try checking Optimize for unreliable recordings.

Tip

For a LIMITED TIME, we are compensating per hour for game data under these circumstances:

  • You must be 18 years old, or older.
  • You must play a game on the games list
  • You must only play PvE or Co-op PvE (ie: gameplay against in-game, non-human opponents). We do not allow PvP recordings.
  • You must upload least 20 hours of footage.
  • You must record active and human play. No camping, bots, idling, etc.
  • You must use Wise or Fiverr. We do not use other payment providers. If you need help setting up ask our Discord community.
    • Due to legal and technical restrictions, residents of MENA countries, .PK, or .VN must use Fiverr.

Troubleshooting

Software known to interfere with OWL Control:

  • MSI Afterburner - Causes problems with recording. Use Steam Overlay with performance stats instead.
  • RivaTuner Statistics Server - Often installed with MSI Afterburner. Sometimes causes conflicts.
  • Antivirus Software - OWL Control is NOT malware. If you experience problems, you are safe to lower antivirus on OWL Control while problem solving.

If you run into other difficulties, write down what happened and take screenshots using Windows' snipper tool, then speak to us on our Discord server or open an issue on GitHub.

Note

You may get an .invalid recording that is marked as Too Long, is longer than 10 minutes, or larger than 150-200MB. If this happens, Please speak to us on our Discord server or open an issue on GitHub.

Contributing to AI Research

By using OWL Control, you're helping to:

  • Train AI agents to understand and play games
  • Develop better spatial comprehension for AI systems
  • Build open datasets for the scientific research community
  • Advance the field of AI and machine learning

Scrubbed and filtered data will be made publicly available for research purposes.

For Developers

OWL Control is open source!

If you're interested in the technical details or want to contribute:

Building from Source

Using powershell:

  1. Install cargo

  2. Clone the repo

git clone https://github.com/Wayfarer-Labs/owl-control.git
cd owl-control
  1. Build the application to create the target directory
cargo build
  1. Install cargo-obs-build
cargo install cargo-obs-build
  1. Install the dependencies
cargo obs-build --out-dir target\x86_64-pc-windows-msvc\debug
  1. Run OWL Control with:
cargo run

To build a production-ready release with an installer:

  • Install NSIS to default location

  • Run the build script

build-resources\scripts\build.ps1

If you'd like to collaborate on this project, please visit our contributor guidelines.

Need Help? Where to Go
🐛 Issues or Bugs? Report them on our GitHub Issues page
Questions? Visit our GitHub Issues page
💬Discord Discord Community

OWL Control is a project by Wayfarer Labs

Building open datasets for AI research


2025 Wayfarer Labs
Trademarks `` copyright respective owners where indicated .

About

OWL Control is a desktop application that records gameplay footage and input data from video games to create open-source datasets for AI research.

Resources

License

Contributing

Stars

Watchers

Forks

Packages

No packages published