PixelFlow

Cross-platform AI video enhancement CLI for video upscaling, image upscaling, frame interpolation, and video restoration. PixelFlow uses Real-ESRGAN, RIFE, and FFmpeg to improve video quality, increase video resolution, generate high-FPS video, and automate AI-powered video processing workflows on macOS, Linux, and Windows.

Upscale. Interpolate. Enhance.

AI Video Upscaling & Frame Interpolation

PixelFlow is an open-source command-line tool for:

• AI video upscaling
• AI image upscaling
• Video frame interpolation
• 30 FPS to 60 FPS conversion
• 60 FPS to 120 FPS conversion
• Video enhancement and restoration
• Batch video processing
• Automated video workflows

Powered by:

• Real-ESRGAN for image and video upscaling
• RIFE for frame interpolation and FPS boosting
• FFmpeg for video decoding, encoding, and media processing

Ideal for content creators, developers, video engineers, AI enthusiasts, and anyone looking to enhance videos using open-source AI models.

Install

PixelFlow is a Python 3.12+ package, published on PyPI as pixelflow-cli (the import package and the command are both pixelflow). For everyday use, install it as an isolated CLI with uv or pipx:

uv tool install pixelflow-cli   # or: pipx install pixelflow-cli

You can also run it once without installing:

uvx pixelflow-cli doctor

Or, inside an existing virtual environment, with pip:

pip install pixelflow-cli

Then provision the external tools (FFmpeg, Real-ESRGAN, RIFE) for your platform:

pixelflow init
pixelflow doctor   # verify everything is ready

doctor prints a report like:

FFmpeg          OK
FFprobe         OK
Real-ESRGAN     OK
RIFE            OK
Backend         OK    ncnn
Platform        OK    macOS arm64
Temp directory  OK

Usage

Common use cases include:

• Upscaling 720p videos to 1080p or 4K
• Converting 24 FPS, 30 FPS, or 60 FPS videos to higher frame rates
• Enhancing anime, gameplay recordings, screen recordings, and archived footage
• Automating video enhancement pipelines from the terminal

# Upscale a single image
pixelflow upscale-image input.png -o output.png

# Upscale a video 2× (audio preserved)
pixelflow upscale-video input.mp4 -o output.mp4 --scale 2

# Interpolate to 60 fps
pixelflow interpolate input.mp4 -o output.mp4 --fps 60

# Upscale and interpolate in one pass
pixelflow enhance-video input.mp4 -o output.mp4 --scale 2 --fps 60

State (downloaded binaries, models, cache) lives in an OS-appropriate application directory; override it with the PIXELFLOW_HOME environment variable. See docs/SPEC.md for the full specification.

Development

# Create an environment and install with dev extras
uv sync --extra dev                       # or: python -m venv .venv && source .venv/bin/activate && pip install -e ".[dev]"

# Quality gate
ruff check .
black --check .
pytest

Packaging & publishing

PyPI (pipx / pip)

The package builds with hatchling into a single platform-independent wheel (the heavy AI binaries are fetched at runtime by pixelflow init, not bundled).

Automated (recommended). Pushing a v* tag triggers .github/workflows/publish.yml, which builds the distributions and publishes them via PyPI Trusted Publishing (OIDC — no API token stored in the repo). One-time setup: add a trusted publisher on PyPI for this repo + the publish.yml workflow and a pypi environment.

# Bump version in pyproject.toml + update CHANGELOG.md, then:
git tag -a v0.1.1 -m "PixelFlow 0.1.1" && git push origin v0.1.1

Manual. To build and upload yourself:

rm -rf dist/
python -m build              # produces dist/*.tar.gz and dist/*.whl
python -m twine check dist/*
python -m twine upload dist/* # username: __token__, password: <pypi token>

Standalone binaries (PyInstaller)

Produce a self-contained executable per platform:

pyinstaller --onefile --name pixelflow pixelflow/cli.py

Build on each target OS (cross-compilation is not supported) to produce:

pixelflow-macos-arm64
pixelflow-linux-x64
pixelflow-windows-x64.exe

Homebrew (macOS / Linux)

Publish a formula to a tap that downloads the release tarball and installs the pixelflow entry point:

class Pixelflow < Formula
  desc "Cross-platform AI video enhancement CLI"
  homepage "https://github.com/PixelPaw-Labs/PixelFlow"
  url "https://files.pythonhosted.org/.../pixelflow-0.1.0.tar.gz"
  sha256 "<sha256>"
  depends_on "python@3.12"
  # ... resource/install blocks ...
end

Then brew install pixelflow/tap/pixelflow.

Winget (Windows)

Submit a manifest to microsoft/winget-pkgs referencing the PyInstaller .exe release asset, then winget install PixelFlow.PixelFlow.

License

PixelFlow is an open-source AI video enhancement project focused on video upscaling, image upscaling, frame interpolation, and cross-platform video processing.

See LICENSE.