Google released new JPEG compressor "Guetzli" - 35% better compression
 

https://research.googleblog.com/2017...urce-jpeg.html
https://github.com/google/guetzli

You either save BW or gain quality. The choice is yours :pimp

Bump for a business post.

unless I misses something, these days transfer rates are so high & drives are so cheap, not really needed, I´d rather not compress my images, quality´s king :2 cents:

This. Nailed it on the head.

Yeah, let's invest more and more money on our end to constantly improve our camera gear and image quality - so we can compress it more and save the cheap ass punters 30% file size.

No thanks.

I agree kind of. Although on site optimization and speed play a part also, but as you said these days bandwidth is cheap and people have a least half decent internet.

:thumbsup

Image load time, which is affected by filesize, is considered by Google as a factor when evaluating pages, specially on mobile.

So it's worth considering for any free/promotional areas.

/thread :2 cents:

Still, very exiting!! :)

Bingo!

I was just going to post that 35% isn't a big deal to us but it is to Google!

Future algo will favor Guetzli encoded images 100% guaranteed :thumbsup

pretty cool, but where was this 2 decades ago when image galleries were popular, everyone was on dialup, and bandwidth cost an arm and a leg? :error

exactly - because on mobile it is not only the loading time - it is the bandwidth limit on most mobile contracts. as far as I know there is no "unlimited" mobile use yet (at least not in Europe - I don´t know if thats existing in the States).

i wait for the day they can do it with gifs or when the first png animation comes up.
but 35% on jpgs is also a help especially on designs and thumbs it can make a HUGE difference.

They are lubing us up with the image algo then gonna stick it all in with their new video algo :1orglaugh

absolutely agreeing with you. Mobile costs are heavier & for promotion or fhg, this could be an intelligent use of this new optimizer :thumbsup

How's this compare to https://github.com/tjko/jpegoptim ?

I'd run a benchmark against jpegoptim, but I can't for the life of me get this to compile or work under CentOS

Alright to save anyone some time, this is impossible to make work under Centos 6 (which in my experience of you are running on your servers) because it's not possible to update glibc to anything newer then 2.12 (and this requires 2.14).

I managed to get this running under windows. It's completely and utterly useless for the needs of the porn industry. It's been processing a 7mbyte jpeg for around 17 minutes now (verbose output is on so it's definitely doing something).

Jpegoptim did it in 1 second:
20170227_113739_HDR.jpg 5312x2988 24bit N Exif JFIF [OK] 7100008 --> 3306762 bytes (53.43%), optimized.

Alright. It finished...

JPEGOPTIM: 7100008 bytes -> 3306762 bytes (less then 1 second)
GUETZLI: 7100008 bytes -> 4652672 bytes (approx 52 minutes)

something doesn't sound right, can you upload the file you tried and exact command you used? One detail you may have overlooked is that Guetzli is supposed to encode uncompressed original -> jpeg, not jpeg->jpeg like jpegoptim does...

I know it doesn't sound right. I didn't think it sounded right either, but I posted up on the github issues page and it's exactly what is expected. It just says high quality images that haven't been compressed by another compressor, not that they cannot be jpeg. Their sample image is a PNG. It also uses an absurd amount of ram.

https://github.com/google/guetzli/issues/96
https://github.com/google/guetzli/issues/50

As it sits, this utility is completely and utterly useless unless you want to devote literally 24 hours of processing time to (very poorly) compressing a single FHG.

Interesting but such requirements and hardware usage for long minutes to get one pic done? That's a tad insane.