Shrink PNG and JPEG files losslessly. The pixels come out identical, only the file gets smaller. Runs on any Linux distribution through Flathub or the Snap Store.
Flathub, for any distribution:
flatpak install flathub com.github.gijsgoudzwaard.image-optimizer
Snap Store, for Ubuntu and anywhere else snaps run:
sudo snap install image-optimizer
elementary OS AppCenter. Note that this channel may lag behind the releases above.
Both the Flathub and Snap builds are published for amd64 and arm64, and both bundle the two optimizers, so there is nothing else to install.
PNG and JPEG. Drop them in and they are rewritten in place, so the file you selected is the file that gets smaller.
Lossless means the pixels come out identical. Nothing is re-encoded at a lower quality, so there is no visible trade-off, and equally no dramatic saving: what comes off is whatever the encoder left on the table.
How much that is depends entirely on the image. Some measurements:
| image | before | after | saved |
|---|---|---|---|
| screenshot, PNG | 24,441 | 19,890 | 18.6% |
| screenshot, PNG | 16,840 | 14,956 | 11.2% |
| exported JPEG | 11,261 | 11,049 | 1.9% |
| 3000x2000 photo, JPEG | 2,535,780 | 2,469,183 | 2.6% |
Screenshots and flat interface images tend to give the most, because their encoders usually settle for the first workable compression. Photos out of a camera or an export dialog are already close to optimal, so a few percent is a good result there. An image that is already optimal comes back unchanged, which is the honest answer rather than a failure.
Two details worth knowing:
- Comments, Exif, IPTC and XMP are stripped, which is part of the saving. The ICC colour profile is deliberately kept, because dropping it makes a wide-gamut image render as sRGB afterwards.
- Modification times are preserved, so optimizing a folder does not reshuffle a photo library sorted by date.
To build you need GTK 4.12 or newer, GLib, a C compiler, Vala, Meson, Ninja,
msgfmt for the translations and update-desktop-database for the install
step. On Ubuntu 24.04 and other Debian derivatives that is:
sudo apt install build-essential meson ninja-build valac gettext \
desktop-file-utils libgtk-4-dev libglib2.0-dev
To run, the app needs the two optimizers it drives. They are not needed to build, but without them nothing gets compressed:
sudo apt install jpegoptim optipng
Two more are optional. Install them before configuring, because Meson looks them up once at that point:
sudo apt install appstream xvfb xdotool
appstream adds the MetaInfo validation to ninja test, which otherwise runs
one test instead of two. xvfb and xdotool are for the scripts under
Tests.
Simply clone this repo, then:
Run meson setup build to configure the build environment and ninja -C build test to build and run the automated tests
meson setup build --prefix=/usr
ninja -C build test
To install, use ninja install, then execute with com.github.gijsgoudzwaard.image-optimizer
sudo ninja -C build install
com.github.gijsgoudzwaard.image-optimizer
ninja test covers the desktop and MetaInfo files. Two more scripts run the
real binary, because a build that compiles is not the same as an app that
works: the GTK4 port of the image list once did the former without the latter.
Install into a staging tree first, then point them at the binary:
meson setup build --prefix=/usr
ninja -C build
DESTDIR="$PWD/dest" ninja -C build install
.github/scripts/smoke-test.sh dest/usr/bin/com.github.gijsgoudzwaard.image-optimizer
.github/scripts/regression-test.sh dest/usr/bin/com.github.gijsgoudzwaard.image-optimizer
smoke-test.sh starts the app on a virtual display, hands it a PNG and a JPEG,
and requires them to come back smaller with nothing logged.
regression-test.sh covers what has actually broken before: awkward filenames,
unreadable files, whole batches being skipped, parallel output matching
sequential, a single core machine, and Ctrl+Q. Both need xvfb, and the
regression suite also needs xdotool.
Both run in CI on amd64 and arm64, so a pull request gets the same answer you do locally.
Released under the MIT License.
Image Optimizer is free and open source. If you find it useful, you can support its development with a one-off or recurring donation:
Contributions are just as welcome. Bug reports, translations and pull requests all help. See the issue tracker to get started.