Guetzli is the new JPEG encoder developed and released as open source by Google. It is designed to produce ~35% smaller files, without compromising the picture quality. We, at ShortPixel, had recently our own update: ShortPixel Glossy optimization, the lossy new algorithm for photographers.
Testing Guetzli algorithm was mandatory for us. When you are in the image compression business, you need to get your hands on any brand new innovation and see what it can do.
How does Guetzli compare with ShortPixel’s new Glossy image optimization algorithm in terms of quality and performance? Can we use it alongside our algorithms? Does it have practical value for our users?
Guetzli was tested on a quad-core Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz processor. We used Guetzli setup with two parameters: 84 and 95. 84 is the minimum quality allowed by Guetzli, and 95 is the default setting (in this case, a bigger number means better quality and larger files).
As we mentioned before, Glossy is our new image optimization algorithm for photographers. The images optimized this way look brilliant while being considerably smaller than their Lossless version. Glossy optimization algorithm runs on ShortPixel cloud servers, where your images are also optimized.
The images compressed in this test are from the amazing Unsplash website. You can find all files, both originals and processed, in the attached archive. At the end of this article, you can also see the complete image credits list.
Guetzli VS ShortPixel: test results
Image size improvement, compression speed and picture quality – these are the most important factors we gathered for this comparison.
File size improvement
ShortPixel Glossy produces smaller images than Guetzli at 95 setting for all tested images, and it has an overall improvement comparable with Guetzli 84.
|Original||Guetzli 95||Guetzli 84||ShortPixel Glossy|
|Img-1||5 MB||1.7 MB||1.1 MB||804.6 kB|
|Img-2||9.3 MB||2.5 MB||1.2 MB||1.5 MB|
|Img-3||1.4 MB||485.1 kB||250.7 kB||97.1 kB|
|Img-4||5.6 MB||1.5 MB||778 kB||1.4 MB|
|Img-5||35.5 MB||10 MB||6.1 MB||6.5 MB|
Speed performance is Guetzli’s weakest point. Google encoder needs several minutes to process the images, while ShortPixel takes seconds to optimize the same files.
|Img-1||9m 01s||9m 40s||8.7s|
|Img-2||9m 29s||6m 30s||6.5s|
|Img-3||7m 56s||4m 53s||3.2s|
|Img-4||7m 02s||7m 45||6.4s|
|Img-5||31m 33s||25m 47||26.8s|
Our users are not scientists, they don’t use image analyzing software, they simply look at images and decide for themselves if they look good or not.
Look at these 100% cropped images and see the differences between Guetzli processed files and the images optimized with ShortPixel Glossy. Remember that WordPress has its own compressing settings and the best way to compare the images optimized with different algorithms is by looking at the full sized pictures found in the attached archive.
Google has a lot of talent resources, and it is not surprising that their algorithms are smart. Guetzli produces good results.
Picture quality is really good when the encoder is setup at 95, but files are pretty big. Changing the option to 84 brings a decrease of picture quality as you could see in the above examples.
Images optimized with ShortPixel Glossy algorithm have a picture quality close to Guetzli-95, but with a pretty good quality-compression ratio. Remember that we recommend ShortPixel Lossy optimization if you want Google PagesSpeed Insights improved results.
The biggest issue of Guetzli in our opinion is the incredible amount of time it takes to deliver the results. You need a computer with a lot of power to run it, and it is very slow. Developers could tweak its parameters, but we are not sure if Guetzli is a practical solution for regular users.
We are not so happy with this conclusion, because we considered useing Guetzli alongside our algorithms. Currently that is not feasible: it takes too much time and processing power to deliver results that are comparable (if not worse) that those delivered by ShortPixel’s Glossy.
Get your image optimization tool of choice!
Finally, you are the jury, and you decide which tool is the best choice for you.
Source: Andrew Phillips
Source: Miguel Mateo
Source: Mitchell Hollander
Source: Robert Zunikoff