Borrowing coding styles and conventions from scikit-image and scikit-learn, scikit-video is a Python module for video processing built on top of scipy, numpy, and FFmpeg.

This project is distributed under the 3-clause BSD.

Visit the documentation at http://www.scikit-video.org

Requirements:

• FFmpeg >= 2.8 on the system PATH (primary tested backend). The libav/avconv backend is retained in the codebase for compatibility but is not validated by the test suite.
• Python >= 3.10
• numpy >= 1.22
• scipy >= 1.9
• Pillow >= 9.0
• opencv-python-headless >= 4.5 (required by skvideo.measure)
• mediainfo (optional)

Installation from PyPI:

pip install scikit-video

Installation from source (GitHub):

git clone https://github.com/scikit-video/scikit-video.git
cd scikit-video
pip install .

For development (editable install with test dependencies):

pip install -e ".[test]"

If you installed scikit-video prior to version 1.1.10, you may have an import conflict with the older sk-video package. Run the following to fix it:

pip uninstall sk-video

Then check that skvideo resolves to the expected location:

import skvideo
print(skvideo.__file__)

scikit-video is actively maintained again. The 1.1.12–1.1.15 line modernizes the package for current Python (3.10–3.13), NumPy 2.x, and SciPy, with no breaking API changes; code that worked with 1.1.11 should continue to work unchanged. Highlights:

• Correctness completion (1.1.15). Remaining user-facing assert validation now raises real exceptions (so it survives python -O); ssim rejects sub-window-size frames instead of returning NaN; last Python-2 xrange removed.
• Non-local I/O (1.1.14). vread / vreader / vwrite and the FFmpegReader / FFmpegWriter constructors now accept file paths, URL strings (http://, https://, rtsp://, ...), and file-like objects (io.BytesIO) interchangeably (issues #117, #113, #81). A UserWarning is emitted if a URL scheme isn't compiled into the local FFmpeg build.
• Python-3 correctness pass. Fixed several public functions that crashed on modern Python (e.g. measure.Li3DDCT_features, utils.canny, motion.globalEdgeMotion, from skvideo import *), replaced assert-based input validation with real exceptions (so it survives python -O), and made probe failures warn instead of silently returning empty metadata.
• Earlier in the 1.1.12–1.1.13 line: pyproject.toml packaging (restores pip install on modern Python/NumPy), pathlib.Path support everywhere, audio passthrough (audiosrc=) and multi-stream ffprobe in the writer, repeated-flag dict values, and a windowed-read start_frame argument.

See CHANGELOG.rst for the complete per-version history.

• Spatial-Temporal filtering helper functions
• Speedup routines (using cython and/or opencl)
• More ffmpeg/avconv interfacing
• Add additional algorithms and maintain more comprehensive benchmarks

See CONTRIBUTING.rst for setup instructions.

After installing with the [test] extra, run the suite with:

pytest -v skvideo/tests

Tests that require FFmpeg or libav are automatically skipped if the corresponding binary is not on the system PATH.

Copyright 2015-2026, scikit-video developers (BSD license).