The Stevens Axiom
COVID Lockdown
the cause of devil tumors
Aurora from Southern Tasmania
European wasps in Tasmania
tranquil Tasmania

19 March 2017

The Net Just Got More Interesting Thanks To Google Guetzli

Graphic designers will tell you the big trade-off was always between image quality and web page load times. You really can't have both. There is javascript code that makes images load as you scroll down but it's mostly used on image galleries. Guetzli has solved a big problem for designers.

Explore the.....wood?

At the end of 2016 Google had a breakthrough with the Guetzli image compressor. Within 24 hours of finding it I was using it.
I downloaded the program from Github and compiled it on an Ubuntu system. It is available for Macs and Windows.
It's console-based, but I guarantee it won't be long before it has a GUI front end. For now, I recommend a command line terminal with copy-n-paste. That will make it a lot easier to process images.
I usually aim for about 150kb image size for the net. I use Gimp but all graphics programs will let you export to JPG at any quality you want.

Guetzli can halve the size of an image and still make it look good. That means I could export to about 200 to 300kbs and then let the Guetzli algorithms crunch it down to something light and fast.

Guetzli makes a folder called 'guetzli-master'. I made a subfolder called 'test' to put my images in.
CD into 'guetzli-master' from your terminal and paste-in this command:
./bin/Release/guetzli --quality 90 --verbose ./test/Image.JPG ./test/Image-guetzli.JPG
It takes a minute or two to process an image and you can watch the matrix-like code do it's thing. The default quality setting is 90% of original but it's recommend not to go below 84%.
For me the command line work flow is easy to use. I check the image quality and size in the test folder and change Gimp and Guetzli settings as needed, deleting as I went along until I had what I was aiming for.
I give Google 10 out of 10 for Guetzli. Designers have waited ages for this and it could even be further improved or inspire new image compression algorithms.

No comments: