The problem with these type of conversions is that most tools don't work properly (rasterize vector formats, etc).
But there seems to be a way to do it correctly with most Windows formats. That's how it's done.
Say you have .doc or .ppt file, from which you need to take figures for LaTeX. You open it with OpenOffice, and save it in OO's format, then unzip the resulting file (it's zip format), and then open the necessary files (usually .wmf or .emf) in "Pictures" directory. Then you export each figure/picture as PDF.
The problem is that OpenOffice exports full page PDF (even if you say it to export the selection only), and one needs small eps/pdf.
So the way to get them is: open the PDF file in Acrobat (full Acrobat), and save as EPS. The EPS is still full-page size, but is still correct (vectors, etc).
So next you use something like the following script on the EPS - what it does is run eps2eps on it (the output eps is garbage, but has correct BoundingBox), and substitute the good BoundingBox in the EPS, and then also runs epstopdf on the good eps, to get good embedded pdf (epstopdf preserves vectors and fonts).
The catch is that you need full Acrobat, which you need to buy from Adobe. :( And it costs a lot of money. :( At least I think it does...
Feb 9, 2005