polytool

One CLI, 26 everyday utilities. Image format conversion, background removal, video/audio conversion, PDF tooling, OCR, QR codes, hashing, downloads, and more — all behind one binary called pt.

Install

The fastest way (no Python prerequisite — uv brings its own):

# 1. install uv (Windows PowerShell)
irm https://astral.sh/uv/install.ps1 | iex

# 2. install polytool (slim — fast, ~80 MB)
uv tool install polytool

# or, install everything (image/video/PDF/yt-dlp/screenshots/AI background removal)
uv tool install 'polytool[full]'

# macOS / Linux
curl -LsSf https://astral.sh/uv/install.sh | sh
uv tool install 'polytool[full]'

Note for PowerShell users: quote 'polytool[full]' — the brackets are shell glob characters.

After install, both polytool and the short alias pt are on your PATH.

Quick start

pt qr gen "https://github.com/k6w/polytool" -o qr.png
pt img convert photo.heic -o photo.jpg
pt img bg-remove portrait.jpg                  # transparent PNG
pt vid gif clip.mp4                            # animated gif
pt vid extract-audio movie.mp4                 # mp3
pt pdf merge a.pdf b.pdf -o merged.pdf
pt pdf to-images doc.pdf --dpi 200
pt dl get 'https://www.youtube.com/watch?v=...'
pt enc hash sha256 release.zip
pt enc base64 encode README.md
pt data convert config.yaml --to json
pt gen password --length 32
pt gen uuid v7
pt color convert "#3366ff"
pt convert unit "100 km" --to mi
pt convert timestamp 1715200000
pt text slugify "Hello, World!"
pt cron explain "0 9 * * MON"
pt net http GET https://api.github.com
pt file dedupe ./Downloads
pt file archive ./project -o project.7z

Run pt --help to see all groups, or pt <group> --help for any group's verbs (each verb's help has runnable Examples).

Documentation

Full reference for every verb, argument, and option lives in docs/:

• Install guide — slim vs full, all extras, Windows quoting
• Per-group reference — img · vid · pdf · dl · shot · qr · data · enc · text · convert · color · cron · file · gen · net · clip
• Architecture — package layout, lazy imports, error model, how to add a verb
• Troubleshooting — every common error with a fix

Feature reference (all 26 verbs)

Group	Verbs
pt img	convert (png/jpg/webp/avif/heic/svg/bmp/tiff), resize, compress, bg-remove, watermark, ascii, exif, palette, ocr
pt vid	convert, trim, extract-audio, gif
pt pdf	merge, split, compress, extract-text, to-images, from-images, ocr
pt dl	get (yt-dlp wrapper, audio-only mode), info
pt data	convert (json↔yaml↔toml↔csv↔xml), pretty, validate
pt enc	hash (md5/sha1/sha256/sha512/blake2b/xxhash), base64, url, html, jwt-decode, jwt-verify
pt qr	gen (png/svg/pdf/eps/terminal), wifi (Wi-Fi join code), decode
pt gen	password, uuid (v1/v3/v4/v5/v7), lorem
pt file	rename, dedupe, bigfiles, organize, archive (zip/tar/tar.gz/7z)
pt net	port-check, ip-info, http
pt clip	copy, paste
pt shot	screen, web (Playwright), install
pt color	convert (hex/rgb/hsl/hsv/cmyk)
pt convert	unit, timestamp, base
pt text	diff, wc, case (snake/kebab/camel/pascal/...), slugify, md-to-html, md-preview
pt cron	explain, next

Slim vs full install

The default install ships only lightweight utilities (data, enc, gen, color, convert, text, qr-gen, cron, net, clip, file). Heavy/optional features are grouped behind extras so you only pull what you need:

Extra	Adds	Approx. size
[img]	Pillow + HEIC/AVIF/SVG plugins, EXIF, palette, ASCII, watermark	~50 MB
[vid]	ffmpeg-python + bundled ffmpeg (auto-downloaded)	~70 MB on first use
[pdf]	pypdf, pikepdf, pdfplumber, PyMuPDF	~80 MB
[dl]	yt-dlp	~10 MB
[shot]	mss, Playwright (Chromium installed via pt shot install)	~150 MB
[ai]	rembg + ONNX runtime (170 MB U2-Net model on first run)	~250 MB
[ocr]	pytesseract / easyocr (PyTorch)	~1 GB if easyocr
[qr-decode]	pyzbar (system libzbar on Linux)	~5 MB
[archive]	py7zr	~5 MB
[full]	all of the above	~1.5 GB

Verbs that need an extra you haven't installed print a friendly hint:

This command needs the 'ai' extra.
Install with: uv tool install 'polytool[ai]'

Troubleshooting

pt img convert *.svg fails on Windows

We use resvg-py (a pre-built Rust wheel) instead of cairosvg precisely to avoid the Cairo DLL pain. Reinstall the [img] extra to pull it in.

pt vid wants ffmpeg

We prefer a system ffmpeg if it's on PATH, and fall back to imageio-ffmpeg's bundled binary (auto-downloads ~70 MB on first call). Install ffmpeg system-wide with winget install ffmpeg / brew install ffmpeg / apt install ffmpeg to skip the bundled download.

pt img bg-remove is downloading something

First run pulls the U2-Net ONNX model (~170 MB) into ~/.u2net/. Set U2NET_HOME to relocate.

pt qr decode says pyzbar missing

On Linux, install the system library: sudo apt-get install libzbar0 (Debian/Ubuntu) or sudo dnf install zbar (Fedora). Then reinstall [qr-decode].

pt shot web fails on first run

Chromium isn't installed yet. Run pt shot install (which is shorthand for python -m playwright install chromium).

pt img ocr says Tesseract not found

The default tesseract engine needs the system Tesseract binary. Either install it (winget install --id UB-Mannheim.TesseractOCR on Windows, brew install tesseract, apt install tesseract-ocr) or pass --engine easyocr to use a self-contained PyTorch OCR.

Brackets in uv tool install 'polytool[full]'

PowerShell parses unquoted brackets — always wrap the spec in single quotes.

Contributing

git clone https://github.com/k6w/polytool
cd polytool
uv sync --extra dev --extra img --extra vid --extra pdf --extra dl --extra shot --extra archive --extra qr-decode
uv run pre-commit install
uv run pytest
uv run ruff check .

Conventional Commits required (feat:, fix:, chore:, docs:, test:, ci:, build:).

Each user-facing verb must:

• Include a runnable Examples: block in its --help.
• Read from path or stdin (-) where it's semantic.
• Default --output to next-to-source for files, stdout for text.
• Raise PolytoolError(msg, hint=...) on user errors (rendered as a red panel).
• Ship at least one happy-path and one error-path test.

License

MIT — see LICENSE.