Turning a Paper Into Slides Without Reading It Twice

Split the job in half and only delegate one of them. The mechanical half, finding every figure, copying the files they need, preserving captions, building the frame skeleton, is genuine work that goes wrong by hand at midnight. The judgment half, deciding what the talk argues and in what order, is the talk. Handing over the second is how you end up presenting your paper's table of contents.

Why section order is the wrong order

A paper is written for someone who can reread a paragraph and skip ahead. Related work goes early so a reviewer can situate the contribution. A talk has none of those affordances: the audience gets one pass, at your pace, and cannot go back. So the same content wants a different order, with the claim near the front and the literature reduced to whatever makes the claim legible. That is the reason a deck built straight from \section headings always feels flat.

The mechanical half, delegated

Inventory first, decide second, lift third. The inventory step matters: you cannot choose which evidence is contested from memory of a forty-page document.

What to do with what is left over

Everything you cut has a home: the appendix. Figures that did not earn a slide become backup slides for the questions they answer, which is exactly the pattern a defence needs and a conference talk benefits from. Mark them noframenumbering so the talk still reads as twelve slides.