I have a manuscript that includes most of its figures by using the \input{filename} command to reference a filename.{tex,eps} pair output from gnuplot using the epslatex terminal. I then run lualatex or pdflatex to compile the document; epstopdf converts the EPS files to PDF on the fly; and the figures are included in the document with vector graphics and native TeX fonts.

Is there any way to make this work for a paper submitted to arXiv?

Apparently, arXiv doesn't do any sort of epstopdf conversion on their side, so that will have to be done locally. I can do that. But converting the original EPS file to PDF produces just the graphical parts of the figure (no text), and is thus incomplete. Another possible solution would be to produce plain postscript output from gnuplot and then convert it to PDF. However this is not ideal, since then you do not retain the native TeX fonts in the figure.

Is there an ideal solution to this problem?

1 Answer 1


I don't think it's a problem since they can compile with latex, not pdflatex, and images can be in EPS format.

Anyway, you can use the standalone option for the epslatex terminal: in this way you create a create a .tex file (of course you can insert a custom preamble in order to have the same font of your main document) to be compiled. You'll have a PDF file at this point. You can either convert the PDF to EPS with the command

pdf2ps -eps file.pdf

or keep just the PDF image and force arXiv to use only pdftex with


within the first 5 lines of the preamble of the main LaTeX file.

  • The figures are slightly different when produced with the standalone option and included with \includegraphics{file} but it works. Thanks!
    – Justin
    Commented Oct 2, 2015 at 16:33

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .