Türkçe · English
Rename and organize your screenshots by what's actually in them.
Your screenshots folder is a wall of Screenshot_20261014_184412.png. You can't
find anything in it, because the names say nothing. shotsort reads the text
inside each screenshot with OCR and renames it after what it shows.
$ shotsort suggest ~/Pictures/Screenshots
3 screenshot(s) would be renamed:
Screenshot_2026-10-01.png -> 2026-10-01_quarterly-revenue-report.png
Screenshot_2026-10-02.png -> 2026-10-02_install-python-on-windows.png
Ekran Goruntusu 3.png -> 2026-10-03_toplanti-notlari.png
This was a preview. Run 'shotsort apply' to do it.The date comes from the file; the rest comes from the heading shotsort read
inside the image. Suddenly the folder is searchable.
Needs Tesseract for the OCR (without it, shotsort still organizes by date — it just can't read the text):
# Tesseract: apt install tesseract-ocr / brew install tesseract / choco install tesseract
pip install shotsortOr from source:
git clone https://github.com/ato113/shotsort
cd shotsort
pip install -e .
shotsort --versionshotsort suggest ~/Pictures/Screenshots # preview the renames (nothing moves)
shotsort apply ~/Pictures/Screenshots # do it (asks first)
shotsort undo # reverse the last runUseful flags:
| Flag | Meaning |
|---|---|
-m, --by-month |
File results into YYYY-MM/ folders |
-r, --recursive |
Include subfolders |
-a, --all-images |
Not just files that look like screenshots |
-l, --lang |
OCR languages, +-joined (default tur+eng) |
--no-ocr |
Date-only names, skip OCR |
-y, --yes |
Skip the confirmation on apply |
- Date — from the file's modification time, as
YYYY-MM-DD. - Title — OCR returns the text top-to-bottom, so the first substantial line after the UI chrome (the clock, the battery, menu bars) is almost always the heading. shotsort filters out that noise and takes the first real line.
- Slug — lowercased, Turkish characters folded to ASCII (
Toplantı→toplanti) so the name is safe on every filesystem, trimmed to a few words.
No usable text? The file still gets a clean YYYY-MM-DD_screenshot.png name.
suggestchanges nothing; you always see the plan first.applyrecords every move, andundoputs them all back.- Running twice is safe: a file already in shotsort's
date_titleshape is left alone. - Name collisions never overwrite — a second file gets
-1,-2, and so on.
OCR quality depends on the Tesseract language packs you have installed. The
default is tur+eng; shotsort automatically drops any language that isn't
installed so it never errors, falling back to English. Install more with your
package manager (e.g. tesseract-ocr-tur).
pip install -e ".[dev]"
pytest # 23 tests
ruff check src testsThe naming logic is tested on its own (no OCR needed), and a separate set of tests generates real images with text, runs them through Tesseract, and checks the resulting names — those skip automatically when Tesseract isn't installed.
The interesting, OCR-free core is naming.py — improving how the title is
chosen is the most valuable contribution. Issues and PRs welcome.
AI-assisted contributions are fine as long as you say so in the pull request and can explain what you're submitting.
MIT — see LICENSE.