#216 · Primary category: Computer Vision

DXcam

2026 computer-use-agent computer-vision d3d11 deep-learning fps-game low-latency python pytorch screen-capture screenshot streaming

A Python high-performance screen capture library for Windows using Desktop Duplication API - Updated 2026

Project last updated:03/18/26

GitHub Stars

798

Forks

99

Contributors

5

License

MIT

Why we included this project

DXcam is a Windows screen-capture library that returns frames as NumPy arrays, so it fits naturally into computer-vision work and AI agents that need to watch a live desktop. It is built on the Desktop Duplication API, with a Windows Graphics Capture backend as an alternative, and it handles full-screen exclusive Direct3D apps that tend to break simpler screenshot tools. The library is tuned for low-latency, high-FPS grabbing, and it includes the controls real capture pipelines need: region capture, target-FPS pacing with drift correction, a ring buffer for continuous recording, zero-copy frame views, and one camera instance per monitor. If you are collecting training data or logging game telemetry, DXcam gets you from screen to inference loop without a separate capture server.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category