Common Video Utilities

A small Python library that wraps common video I/O tasks — reading metadata, iterating frames, and writing frames back to a video file — behind a single, consistent interface across four backends: OpenCV, MoviePy, PyAV, and decord.

Features

• Read video metadata (width, height, FPS, frame count, duration) with video_info().
• Stream frames from a video one at a time with video_frames(), without loading the whole file into memory.
• Write frames to a new video file with frames_to_video().
• Create an OpenCV VideoWriter matching the properties of an existing video with video_writer_like().
• Swap between "opencv", "moviepy", "pyav", and "decord" readers/writers per call, so you can pick whichever backend performs best or supports your codec.

Dependencies

pip install opencv-python-headless moviepy decord av

Installation

This library is designed to be used as a git submodule:

git submodule add https://github.com/rendicahya/python_video.git

Available Functions

1. video_info()

Returns a dictionary containing information about a video:

• width
• height
• fps
• n_frames
• duration (in seconds)

Available readers are "opencv" (default), "moviepy", "pyav", and "decord". path can be either a string or a pathlib's PosixPath.

Usage:

frompython_videoimportvideo_infofrompathlibimportPathpath="/path/to/video"# or:# path = Path("/path/to/video")info=video_info(path, reader="opencv")

2. video_frames()

Returns a frame generator from the specified video. Available readers are "opencv" (default), "moviepy", "pyav", and "decord".

Usage:

frompython_videoimportvideo_framespath="/path/to/video"# or:# path = Path("/path/to/video")frames=video_frames(path, reader="opencv")
forframeinframes:
# ...

To grab all frames at once, cast to a list. This should be avoided, though, as it inefficiently loads all frames onto memory:

frames=list(video_frames(path))

3. video_writer_like()

Returns an OpenCV video writer with the same properties as the specified video.

Usage:

frompython_videoimportvideo_writer_likesrc_path="/path/to/src-video.mp4"# or:# src_path = Path("/path/to/src-video.mp4")dst_path="/path/to/dst-video.mp4"# or:# dst_path = Path("/path/to/dst-video.mp4")writer=video_writer_like(src_path, target=dst_path)
forframeinframes:
writer.write(frame)
writer.release()

4. frames_to_video()

Writes image frames to a video file. Available writers are "opencv" (default), "moviepy", and "pyav".

Usage:

path="/path/to/target/video"# or:# path = Path("/path/to/target/video")frames_to_video(frames, path, writer="opencv", fps=30, codec="mp4v")

Combining with video_info() and video_frames() to read frames from a video and write them to another video:

src_path="/path/to/target/video-src.mp4"# or src_path = Path("/path/to/target/video-src.mp4")info=video_info(src_path)
frames=video_frames(src_path)
dst_path="/path/to/target/video-dst.mp4"# or dst_path = Path("/path/to/target/video-dst.mp4")frames_to_video(frames, dst_path, writer="opencv", fps=info["fps"], codec="mp4v")

License

Licensed under the MIT License.