An irreverent idea that has now been set loose upon the world.
If you're not familiar with the *_ebooks concept then the Wikipedia article on the most well known of them should fill you in. fansub_ebooks takes the idea, removes the spam and uses scripts from anime fansubs as its corpus. So you can get wonderful, context-free non-sequiturs such as:
My butt... I mean, who cares?
— Fansub Ebooks (@fansub_ebooks) December 30, 2012
— Fansub Ebooks (@fansub_ebooks) December 29, 2012
It posts every half an hour so it won't clog up your timeline.
If you don't have Twitter or just want to poke around the madness, there is a web interface that allows you to view all of the currently uploaded scripts as well as vote on choice lines. Lines with more votes are more likely to be tweeted next.
How you can help
The current database of lines stands at around 150,000 which is the 300 or so fansub files I happened to have to hand. If you'd like to help and are reasonably technically savy (e.g. able to extract subtitle files from MKVs using mkvtoolnix) then it would be a great help if the range of different series and movies represented was greater. Just name the subtitle file (Advanced Substation Alpha / .ass files only at the moment) the same as the fansub file (e.g. [gg] Chuunibyou Demo Koi ga Shitai! - 01 [5B6EFD1F].mkv would become [gg] Chuunibyou Demo Koi ga Shitai! - 01 [5B6EFD1F].ass), zip them up and e-mail them over to fansub_ebooks at chaostangent dot com. Credits, kudos and favours to any and all help.
Special mention to anyone who can provide such legendary scripts as AnimeJunkies 16th episode of Ghost in the Shell: Standalone Complex, Ashiteruze Baby's corm or similar.
I'm still looking for a suitable Twitter avatar and header image (my five minutes of browsing danbooru's "reading" tag was largely fruitless) so any ideas are welcome. Likewise any thoughts on a good short URL for better linking are encouraged, otherwise I may retrofit the adesi.co domain I have (n.adesi.co is for general links, perhaps f.adesi.co?).
Bug reports and feature requests for the web interface are never unwelcome, see below for possible routes.
At the moment the method for choosing what to tweet next wholly favours lines which have recent positive votes (in the last 24 hours); in theory if there is ever diverse enough voting going on this would cause certain lines to never be tweeted. It does however log which lines it tweets so it will never tweet the same line from the same file twice.
The oddities in some of the lines (e.g. #127987) are largely due to the rough and ready nature of the ASS parser I built in between Christmas drinks. Ideally these would not appear and need to be filtered out as well as potentially removing one character "lines".
There is an undocumented REST API which will eventually become a documented REST API once it has become stable enough for general usage.
Likewise once the functionality has stabilised I'll release all of the code under a Creative Commons license on Github.
An assortment of possible improvements and features should the time and inclination arise:
- Twitter commands e.g. "@fansub_ebooks random" to be tweeted a random line, "@fansub_ebooks #127803" for a specific line etc.
- Story mode - allow lines to be collated into ordered lists via the web interface for ad hoc stories
- "Tweet me" link on web interface for sending a specific line in a tweet
- Search functionality - like looking up dirty words in the dictionary
You just wait, this will be as big as cheese one day!