Open bleuge opened 7 years ago
We're now running https://imgbot.net across https://github.com/libretro-thumbnails
This can be closed :+1:
Wow, interesting, very nice work imgbot, and integrated in github. Any stats about what savings could losslesscompress in imagemagick can achieve? Thanks for this!
The reports are on the commits themselves: https://github.com/libretro-thumbnails/Nintendo_-_Game_Boy_Advance/commit/53f96f4a838bde6deee401c91136a38513c9f360
Thanks!, i've checked some of them, something seems broken:
/NamedSnaps/Spirits Spells (USA).png -- 6.70kb -> 0.00kb (100%) /NamedTitles/Legend of Zelda, The - A Link to the Past Four Swords (Europe) (En,Fr,De,Es,It).png -- 7.54kb -> 0.00kb (100%) /Named_Titles/Kami no Kijutsu - Illusion of the Evil Eyes (Japan).png -- 11.92kb -> 0.00kb (100%) /Named_Snaps/Space Invaders (France).png -- 12.27kb -> 0.00kb (100%) /Named_Snaps/Colin McRae Rally 2.0 (USA) (En,Fr,De).png -- 12.22kb -> 0.00kb (100%) /Named_Snaps/Capcom Classics Mini Mix (USA).png -- 11.04kb -> 0.00kb (100%) /Named_Snaps/Spyro Advance (Japan).png -- 7.83kb -> 0.00kb (100%)
Links?
It's the link you sent me, mate :)
The reports are on the commits themselves: libretro-thumbnails/Nintendo_-_Game_Boy_Advance@53f96f4
looks like it doesn't commit the 100% ones: https://github.com/libretro-thumbnails/Nintendo_-_Game_Boy_Advance/blob/master/Named_Titles/Catz%20(USA%2C%20Europe).png
Hi, I've been working here optimizing the PNGs for size.
Used Fileoptimizer that runs some png optimizers serially, too slow (weeks of processing). But found some good compromise in terms of final size and compression time. There is a small loss in the image, not visible even with zooming and just relative to quantization (first step is pngquant), other things as re-compression, huffmann optimizations or stripping tags are lossless.
As other user said, pngs with alpha channel must be checked, AFAIK this is not important in boxes, snaps and titles as they are simple scans or snapshots.
If there is interest in this I can show the process i've choosen (pngquant, truepng, defluff,...) and could try to code a python script for processin the whole set.