Resizing my website’s pictures with ImageMagick and find(1)

I have noticed that most of the pictures I have uploaded on this website are incredibly large:

$ du -h -d 1

Stupid modern phones and their multi-megapixel cameras! I definitely do not want my minimalist blog to waste your bandwidth, I have to fix this. I could just go through all my pictures and resize them one by one with something like gimp - there are like 10 of them. But that’s boring. Let’s do it with the command line instead!


The obvious question is: how do we even edit a picture with the command line? Luckily, there is a tool for that: ImageMagick. This piece of software can do a ton of things, and I do not use it very often, so I always need to look up what I want to do.

One way to invoke ImageMagick is by calling the convert command, which can take a -resize option, followed by a geometry argument. After checking the online manual, I understood that the command I was looking for was:

$ convert picture.jpg -resize "750>" picture.jpg

which resizes picture.jpg by scaling it down to at most 750px width - preserving the ratio between width and height, and leaving the picture untouched if it is already smaller. The value 750 was chosen after a couple of attempts, it seems a good compromise between quality and size.

And now I just have to do this for all pictures. Of course, running the same command 10 times with a different argument is out of question.


(No “man page reading club” this time, but don’t worry, the series will be back soon.)

A standard UNIX command, find allows you to scan a folder for files with certain properties (for example, a certain name pattern) and perform actions on them (for example, running a command). The OpenBSD and GNU versions of find have some differences, check your local manual page. The commands I use here have been tested on the GNU version, but should be standard.

To look for files and simply print their name, we can use -name:

$ find src -name \*.jpg -or -name \*.png

Notice how I had to escape the *: the strings \*.jpg and \*.png are part of the find command, we don’t want the shell to expand them.

Now we can use -exec to run a command on each of these files, using {} to refer to the file’s name:

$ find src \( -name \*.jpg -or -name \*.png \) -exec convert {} -resize "750>" {} \;

We get a couple of warnings about grayscale images, but whatever. This seems to have worked.

The weight of history

Let’s see if things have improved:

$ du -h -d 1

That’s… a bit disappointing. It looks like large pictures were not the only culprit - all the slides for my math talks are also quite heavy. Too bad, I am not going to change them now. Maybe in a follow-up post :-)

Just like me with my old slides, git also wants to keep a memory of the past. After all, this is what version control system are supposed to do. If you are wondering why my website is a git repository, check out this old blog post of mine. In the end, I don’t really need the ability to revert my website to an older version, so I could just reset the repo from all the history at some point.


ImageMagick is and find are both powerful tools. I am happy with looking up ImageMagick’s syntax everytime, but I definitely want to become more proficient with find.

This was not a hard task, but I have learned something new. I hope you did too :-)