Mode d’emploi
- Copiez la commande et examinez le paquet avant de l’installer dans votre environnement agent.
- Consultez le document original pour les outils, identifiants et environnements requis.
- Décrivez les entrées et le livrable souhaité ; comparez la sortie aux sources.
npx skills add https://github.com/immamdouhaboammar/motion-graphics-skills --skill opencv-fast-rendererAvant de commencer
- L’installation n’inclut ni abonnement au modèle, ni quota API, ni accès aux fichiers.
- Étoiles et installations reflètent l’activité, pas un test indépendant de qualité ou de sécurité.
Rendu depuis le dépôt source en conservant titres, exemples, code, tableaux, liens et images.
OpenCV Fast 2.5D Renderer
A lightweight programmatic motion graphics pipeline built on Python, NumPy, Pillow, and OpenCV, streaming directly into FFmpeg.
Why Headless Python Rendering
Browser-based renderers (such as Puppeteer or Chromium frame captures) provide high design flexibility but require heavy browser runtimes and significant memory overhead. For algorithmic particle effects, procedural data visualization, mathematical curves, or serverless execution, the Python OpenCV pipeline renders up to 5 times faster with zero browser dependencies.
Architecture Pipeline
- Procedural Geometry (NumPy): Generate and interpolate 2.5D vector vertices, Bezier control paths, and camera matrices.
- Typography and Antialiasing (Pillow): Render clean subpixel text layers using TrueType or OpenType font files.
- Compositing and Motion Blur (OpenCV): Blend multi-layer frame buffers with alpha channels, apply directional box blur or Gaussian motion blur, and grade color.
- Direct FFmpeg Pipe: Stream raw RGB frame bytes directly to an FFmpeg subprocess via stdin pipe, avoiding intermediate disk writes.
Streaming Execution Pattern
import subprocess
import cv2
import numpy as np
width, height, fps = 1920, 1080, 60
cmd = [
"ffmpeg", "-y", "-f", "rawvideo", "-vcodec", "rawvideo",
"-s", f"{width}x{height}", "-pix_fmt", "bgr24", "-r", str(fps),
"-i", "-", "-c:v", "libx264", "-pix_fmt", "yuv420p", "-crf", "18", "output.mp4"
]
proc = subprocess.Popen(cmd, stdin=subprocess.PIPE)
for frame_idx in range(fps * 10):
frame = np.zeros((height, width, 3), dtype=np.uint8)
# Render procedural scene
proc.stdin.write(frame.tobytes())
proc.stdin.close()
proc.wait()Further Reading
- references/opencv-compositing.md for alpha blending and layer math.

