Have you looked into what mwoffliner does? https://sourceforge.net/p/kiwix/other/ci/master/tree/mwoffliner/mwoffliner.… Maybe you can even just extract the images from the ZIM files. Nemo

As another suggestion, XOWA (http://gnosygnu.github.io/xowa/) can generate a list of thumbs. It takes about 60 hours to parse the English Wikipedia dump and generate a table of 4.78 million rows with the following columns: * file name * file extension * repo (commons or local) * file width * file height * thumbtime (for video) * page (for djvu / pdf) There's more information in XOWA at home/wiki/Help:Import/Command-line/Thumbs . I can provide more information online or offline if you're interested. If you need the actual thumb files, you can download XOWA databases from https://archive.org/details/Xowa_enwiki_latest . They have about 5 million thumbs within SQLite tables. It should be straightforward to write code to pull the blob from the database and save them to disk. Otherwise, as others have indicated, I know of no MediaWiki way to get this information (via .sql dump file or by api.php). Since XOWA parses wikitext, it can generate the information easily, though the solution is not officially a MediaWiki one. Hope this helps. On Sun, Sep 13, 2015 at 1:43 AM, wp mirror <wpmirrordev(a)gmail.com> wrote: