In dealing with DJVU files on English Wikisource, I've sometimes come across files where in the scans (source djvu) the scan orientation is portrait, but the content on the page suggests it should be in a landscape orientation.
This creates 2 problems.
i) It is not possible to OCR these pages as easily.
ii) It is less easy to confirm the WikiText version of the page matches the printed version.
There are two responses to this..
i) Fix the source file, which is not straightforward on some complex djvu files ( and creates a duplicate version of the file at Commons potentially.)
ii) Provide a means of rotating the image in browser, (and to inform the OCR gadget that it needs to flip or rotate the source image prior to applying the OCR algorithm used.)
This task has been opened as a feature request for the functionality to :
i) Provide a gadget on commons which will rotate a single page inside a DJVU file, and uploaded a patched/corrected djvu file, automagically.
ii) Implement a 'rotate page view' function in the modified editing page provided by the Proofread Page extension.
iii) Provide a means for the OCR gadget to be told a page scan orientation needs to be modified before applying OCR.