🌟 World's Fastest Python Screenshot Library for Windows 🐍
import bettercam
camera = bettercam.create()
camera.grab()
BetterCam is the World's 🌏 Fastest Publicly available Python screenshot library for Windows, boasting 240Hz+ capturing using the Desktop Duplication API 🖥️💨. Born from DXCam, it shines in deep learning pipelines for FPS games, outpacing other Python solutions like python-mss and D3DShot.
BetterCam's superpowers include:
- 🚅 Insanely fast screen capturing (> 240Hz)
- 🎮 Capture from Direct3D exclusive full-screen apps without interruption, even during alt+tab.
- 🔧 Auto-adjusts to scaled / stretched resolutions.
- 🎯 Precise FPS targeting for Video output.
- 👌 Smooth NumPy, OpenCV, PyTorch integration, etc.
💞 Community contributions warmly invited!
pip install bettercam
Note: 🧩 OpenCV is needed by BetterCam for color space conversion. Install it with pip install opencv-python
if not yet available.
Each monitor is paired with a BetterCam
instance.
To get started:
import bettercam
camera = bettercam.create() # Primary monitor's BetterCam instance
For a quick snap, call .grab
:
frame = camera.grab()
frame
is a numpy.ndarray
in the (Height, Width, 3[RGB])
format by default. Note: .grab
may return None
if there's no update since the last .grab
.
To display your screenshot:
from PIL import Image
Image.fromarray(frame).show()
For a specific region, provide the region
parameter with a tuple for the bounding box coordinates:
left, top = (1920 - 640) // 2, (1080 - 640) // 2
right, bottom = left + 640, top + 640
region = (left, top, right, bottom)
frame = camera.grab(region=region) # A 640x640x3 numpy ndarray snapshot
Start and stop screen capture with .start
and .stop
:
camera.start(region=(left, top, right, bottom)) # Capture a region (optional)
camera.is_capturing # True
# ... Your Code
camera.stop()
camera.is_capturing # False
When capturing, grab the latest frame with .get_latest_frame
:
camera.start()
for i in range(1000):
image = camera.get_latest_frame() # Waits for a new frame
camera.stop()
cam1, cam2, cam3 = [bettercam.create(device_idx=d, output_idx=o) for d, o in [(0, 0), (0, 1), (1, 1)]]
img1, img2, img3 = [cam.grab() for cam in (cam1, cam2, cam3)]
To list devices and outputs:
>>> import bettercam
>>> bettercam.device_info()
>>> bettercam.output_info()
Select your color mode when creating a BetterCam instance:
bettercam.create(output_idx=0, output_color="BGRA")
We support "RGB", "RGBA", "BGR", "BGRA", "GRAY" (for grayscale). Right now only numpy.ndarray
shapes are supported: (Height, Width, Channels)
.
Frames go into a fixed-size ring buffer. Customize its max length with max_buffer_len
on creation:
camera = bettercam.create(max_buffer_len=512)
For precise FPS targeting, we use the high-resolution CREATE_WAITABLE_TIMER_HIGH_RESOLUTION
:
camera.start(target_fps=120) # Ideally, not beyond 240Hz.
For constant framerate video recording, use video_mode=True
during .start
:
# Example: Record a 5-second, 120Hz video
camera.start(target_fps=target_fps, video_mode=True)
# ... Video writing code goes here
Call .release
to stop captures and free resources. Manual deletion also possible:
del camera
cam = bettercam.create()
# ... Benchmarking code...
BetterCam Nvidia GPU 🏁 | BetterCam 🏁 | DXCam | python-mss | D3DShot | |
---|---|---|---|---|---|
Avg FPS | 111.667 | 123.667 | 39 | 34.667 | N/A |
Std Dev | 0.889 | 1.778 | 1.333 | 2.222 | N/A |
# ... Sample code to test target FPS ...
Target/Result | BetterCam Nvidia GPU 🏁 | BetterCam 🏁 | DXCam | python-mss | D3DShot |
---|---|---|---|---|---|
120fps | 111.667, 0.889 | 88.333, 2.444 | 36.667, 0.889 | N/A | N/A |
60fps | 60, 0 | 60, 0 | 35, 5.3 | N/A | N/A |
- DXCam: Our origin story.
- D3DShot: Provided foundational ctypes.
- OBS Studio: A treasure trove of knowledge.