I’ve never played/watched/read any Fate/whatever games/anime/manga, but Grand Order’s Mash/Mashu/Matthew Kyrielight/Kyrielite in a blue gingham dress with sweater and glasses pushes all of my buttons. The ensemble even has its own tag on Pixiv: 先輩を殺す私服 (senpai wo korosu shifuku), “clothes which kill senpai” (aka the player). A few people have tried to expand the theme to other characters, but it hasn’t taken off.
Amusingly, I typed the above at work, then went over to the break room to nuke my dinner, and three attractive asian women walked in wearing flirty little sundresses with sweaters, and they all had glasses, too.
In related news, I’ve got Pixivpy working well enough that I can extract metadata for an image based on the ID in its filename, store them in a SQLite DB along with my work-in-progress tag translations, generate HTML to post the selected images, and update my Pixiv bookmarks with whatever I just blogged. That last bit helps improve their recommendation system, another way of weeding out the chaff.
Senpai-killing body pillow (NSFW, of course)
Trying to find work-safe pictures of Ikumi Hisamatsu anywhere is pretty much a lost cause. Wearing very little very well is her job, after all.
For some reason, it’s difficult to find not-entirely-unsafe images combining the 猫耳 and 極上の乳 tags, so most of this set ended up hidden in the NSFW section.
This is an experiment in using only positive tags to select images, as opposed to the laundry list of exclusions Steven had to use to weed out the mountains of hideous crap on Gelbooru. Having a paid Pixiv membership allows me to sort by likes, which tends to eliminate that sort of thing, although the Pixiv community has more affection for certain things than I do…
No matter how hard they try (and they really, really do), the Hello!Project Costume Designers can’t make Maimi ugly.
Other not-uglies below, with the usual to-be-named-later.
Need a little touchup on the script still, but I merged together tag translations from two sources (1, 2) and cleaned up a lot of the cruft, ending up with 8,238 tags and (mostly-reasonable) translations. Next will probably be adding a uniq() function to cover duplicates created by the translations.
Of course, the moment I ran the script on this small set of images, it immediately barfed out 107 tags that weren’t on the list…
(Seriously, though, in all these years, no one had added translations for horn, saké, bento, white_hair, black_bra, long_hair, frilly_panties, or high_quality_panties? Most of the rest were series-related, at least)
Working on upgrading my posting script to extract keywords and proper URLs from Pixiv downloads. PixivPy mostly works, although Python is my least-favorite language to hack about in. I haven’t found a good mapping table to create English versions of most tags, even though I know Pixiv has one. The API is kind of goofy, since it’s straightforward REST, but relies on pretending you’re the official Android/iOS client to get authentication to work. Documentation is lacking in two languages.
“Any thing done for the first time unleashes a cheesecake.”
The power went out at home (~4,000 people affected…) while I was prepping this set, so no names for now.
[Update: filled in most of the first half; annoyed that Google just returns “girl” for the first one…]