Skip to content
/ zbl Public

real-time window capture library based on D3D11 and Windows.Graphics.Capture

License

Notifications You must be signed in to change notification settings

modelflat/zbl

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

45 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

zbl

PyPI version

zbl is a Rust and Python library for screen/window capturing on Windows. It provides an interface to Windows.Graphics.Capture API with a focus on simplifying integrating computer vision applications for Windows Desktop apps.

This library is in early development stage. This means that it's only verified to work for a 'happy path' scenarios - beware of bugs!

Python

Installation

pip install zbl

Alternatively, you can install suitable wheel from releases page.

Usage

from zbl import Capture

with Capture(window_name='visual studio code') as cap:
    frame = next(cap.frames())
    print(frame.shape)

The snippet above will capture a window which title contains the string visual studio code, take one frame (which is represented as a numpy array) and print its shape.

See Capture constructor for more options. It is possible to capture the entire screen using display_id argument, for example.

To run an example using OpenCV's highgui:

  1. Install opencv-python
  2. Run python -m zbl --display-id 0

Rust

See examples. Note: if you are getting OpenCV build errors when building the example, check out how to build OpenCV rust bindings.

Comparison to mss / pyautogui

Both are very slow at the time of writing. mss tops at 30-50 fps in a tight loop, pyautogui is even slower than that. zbl is able to capture an order of magnitude faster (at 500-700 fps). This allows a lot more time for the actual processing.

Plans

  • (in progress) Integration with GpuMats & full on-GPU processing

Credits

zbl is heavily inspired by screenshot-rs.

About

real-time window capture library based on D3D11 and Windows.Graphics.Capture

Resources

License

Stars

Watchers

Forks

Packages

No packages published