How to turn a research paper into slides

A paper and a talk do different jobs. The paper has to be complete, so a reader can check every step. The talk has to be clear, so a room of people who can't scroll back leaves with one idea. That's why pasting the paper's sections onto slides in order rarely works, and why a 15-minute talk needs far less of the paper than it feels like it should.

Here is a process that works for conference talks, lab meetings and journal clubs, drawn mostly from MIT's Communication Lab guides for engineers and scientists.

1. Know the slot and the audience

Start with the constraints: how long you have, how long questions take, how big the room is and who will be in it. A guide on oral presentations in the International Journal of Surgery Oncology notes that conference talks are usually short, often around 8 to 10 minutes, and suggests about 1 minute per slide as a rough guide (Gundogan et al.).

Treat that as a starting point, not a rule. MIT's Nuclear Science and Engineering Communication Lab advises against fixating on slide count at all, because complex content may need animation steps that throw the count off. Plan around time and what the audience can absorb instead (MIT NSE Communication Lab).

Then think about who is listening. Specialists at a conference want your contribution and how it compares with prior work. A mixed department audience needs more background. Your own lab wants the details you'd normally cut.

2. Write your message in one sentence

Before opening any slide software, write the single sentence you want people to remember. The NSE guide puts it simply: if someone thinks back on your talk and remembers one thing, what should it be?

MIT's EECS Communication Lab quotes the mathematician Gian-Carlo Rota on why this matters: "Every lecture should state one main point and repeat it over and over" (MIT EECS Communication Lab). A paper can argue several things. A short talk should argue one.

3. Plan from your results outward

The NSE guide describes a talk as an hourglass: start broadly with a problem your audience cares about, narrow into your work, then widen again to what the results mean. But it recommends planning from the narrow middle. Look at your results first and decide what points they support.

Group your results by the point they make, not by the order you ran the experiments, and aim for 2 to 4 groups that each support your main message. Results that don't support one of those points are candidates to cut, however hard they were to get.

4. Decide what to cut from each section

The right cut depends on the talk's purpose. The EECS guide says that when you're presenting a new result, focus on results and implications, and describe methods at the level of "I imaged the magnetic field" rather than naming the camera. When you want feedback on an experimental strategy, as in a lab meeting, flip it: spend the time on methods and alternatives, and tell the audience what kind of critique you want.

5. Turn each point into a slide title

The NSE guide's final step is to write your motivation, methods, results and impact as a series of sentences, one per slide. Those sentences become the slide titles, and read in order they should tell the whole story.

This is the core of the assertion-evidence approach developed at Penn State: a short sentence headline that states the takeaway, supported by visual evidence instead of a bulleted list. Its authors call PowerPoint's default of a phrase headline over bullets the biggest single mistake in technical talks (Assertion-Evidence Approach). Compare "Results: cell viability" with "Coating A kept 90% of cells alive after 48 hours". The second tells the audience what to look for before they look.

6. Rebuild your figures for a screen

Journal figures are made for a page someone can study. On a slide they're seen once, from across a room, while you talk. So rebuild them rather than pasting them:

• Split multi-panel figures. Show the panel that proves the current point, and move the rest to backup slides.
• Enlarge everything. MIT's AeroAstro Communication Lab advises the largest font the slide allows and never below 16 pt, with contrasting colors for plots with many lines (MIT AeroAstro Communication Lab).
• Introduce the axes before the data. The EECS guide suggests telling the audience what the axes are and what pattern to look for first, and building a figure up one piece at a time, for example simulation, then the expected trend, then the measurements (MIT EECS Communication Lab).
• Keep only the equations you need. MIT's Mechanical Engineering guide recommends using slides for figures and equations with as few words as possible (MIT MechE Communication Lab). Show the governing equation, not the derivation, and say what each term means.

7. End on the impact, not "Questions?"

The NSE guide recommends ending by connecting your results back to the problem the whole audience cares about, rather than on a "Questions?", "Thank you!" or backup slide. Leave that final slide up during Q&A, and keep backup slides after it for the questions you expect: controls, extra results, the full derivation.

8. Rehearse with a clock and an audience

Run the talk out loud at least twice with a timer. Gundogan and colleagues recommend presenting to colleagues and senior researchers first so they can ask the questions you're likely to get (Gundogan et al.). If you run long, cut a result group, not your motivation or your conclusion.

Getting a first draft faster

The slowest part is often the blank deck. You can paste the abstract and your one-sentence message into an AI tool and ask for slide titles, or let a tool like MakeDecks build a first draft straight from the paper's PDF and edit from there. Either way, you still choose the message, check every number against the paper and rebuild the key figures yourself. Our page on turning a research paper into slides shows an example, and the same approach works for any long PDF report.

FAQ

How many slides do I need for a 15-minute talk?

About one slide per minute of speaking time is a common starting point, so roughly 10 to 12 slides if you leave time for questions. Slides that build up step by step take less time each, so plan by minutes rather than slide count.

Should I use bullet points on research slides?

Use them sparingly. The assertion-evidence approach replaces topic headlines and bullets with a one-sentence takeaway and a figure, diagram or image that supports it.

Can I reuse figures from my paper?

Yes, but redraw or crop them for the screen: one panel per point, larger labels, fewer lines and the important result highlighted.

How is a lab meeting talk different from a conference talk?

A conference talk sells one result to outsiders and skips most methods. A lab meeting is a working session, so show methods, raw data and failed attempts, and say what feedback you want.