Extraction or recognition
A Gujarati PDF saved from a Unicode document already holds its text, so converting it is extraction and comes out close to exact. A scanned Gujarati PDF holds pictures of pages, so it is recognised first and inherits recognition’s mistakes. You upload both the same way. It is worth knowing which you have before judging the result.
LMG Arun and the typewriter layout
A great deal of older Gujarati DTP was typed in legacy fonts on the Gujarati typewriter layout, and LMG Arun is among the most downloaded of them. Such a document stores the Latin keys the typist pressed, and only that font draws them as Gujarati, so copying from it gives Latin letters. Our legacy converter has a table for LMG Arun, read glyph by glyph from the font itself. Choose Gujarati — LMG Arun in its script list. A Gujarati reader has not yet checked the table, and the converted document says so.
What is kept
Text, paragraph breaks and reading order are kept. Fonts and precise layout are approximated, since the aim is a document to edit. Gujarati conjuncts are written as standard Unicode sequences, so Shruti or any other Unicode Gujarati font draws them properly.