How I Prepare a Research Talk: Notes on Slides, Scripts, and Delivery
I want to share how I prepare a research presentation from the perspective of a student—specifically, an undergraduate.
To me, a research presentation is not just a slideshow. It is a guided conversation where you take an audience from knowing almost nothing about your work to understanding your main claims and why they matter. The goal is simple: by the end, people should know what you did, why you did it, and why they should trust the result. To reach that goal, I think in three layers: slides give the talk structure, the script gives that structure clear language, and delivery makes the argument feel stable and human.
Before diving into each layer, I should mention that my workflow always follows the same order:
- first I build the skeleton of the slides;
- then I write the script;
- finally I return to the slides to fill in figures, emphasis, and refinements that the script reveals to be necessary.
In short:
- Slides provide structure
- Script gives the structure clear language
- Delivery animates the whole thing
⸻
1. Slides: Why I Start With Logic
When I begin working on slides, I always start with logic.
This is because the slide deck is ultimately a story, and stories fall apart if their underlying reasoning is shaky. A research talk becomes much easier to follow when its logical structure is clean, sequential, and intentional. If I begin with colors, layouts, or visual style instead of reasoning, I often end up decorating an argument that isn’t fully formed. But when the logic comes first, every slide has a purpose, and the design naturally supports the message.
Before opening any software, I outline the entire argument on a blank page: the core problem, what current methods do, why those methods fall short, what my idea is, how it addresses the failure modes, what the results look like, and how those results should be interpreted.
This is just an example—the outline can be much deeper depending on the work—but these are the minimum questions that form the backbone of most research talks.
This outline becomes the structural skeleton. The script I write later grows on top of it like flesh over a frame, and the quality of that script depends directly on how clean the underlying structure is.
⸻
2. Clarity Above All
I design slides with one guiding principle: clarity above all.
A research audience can lose track for many reasons. They may zone out for a moment, get stuck thinking about a previous point, miss a definition, or simply need a second to process a dense idea. Because this happens naturally, each slide must be readable enough that an audience member who momentarily lost the thread can look up and instantly understand where we are.
To achieve this, I rely on four techniques:
First, I use bullet points, but each bullet must be short—sometimes just a phrase. A long bullet defeats the point.
Second, I bold or color only one or two phrases per slide—the absolute core ideas that I want the audience to remember. Over-highlighting reduces attention rather than directing it.
Third, I reduce the amount of content on each slide. The slide should contain a complete claim or a complete piece of evidence, but never more than that. If a slide tries to express too many thoughts, they blur together.
Fourth, I use figures selectively. A good figure does what five sentences cannot: it encapsulates an idea in an instantly digestible form. When I need visual evidence, I sometimes—and I emphasize personally—use GPT’s deep research tools to find relevant papers, examine several sources, and choose a figure that is clear and aligns with the argument I want to support. Whenever I include a figure, I cite its source. Citations not only respect the original researchers but also strengthen the credibility of the talk by grounding claims in real literature.
Good figures are not decoration; they are visual arguments.
⸻
3. Scripts: Short Sentences and Visible Logic
Once the skeleton slides are drafted, I write the script. This is where I find myself thinking the hardest, because the script must balance precision, tone, pacing, and logic, all while remaining speakable.
I keep a strict stylistic rule: short sentences only.
This is not a stylistic quirk—it’s practical. Research talks unfold in real time, and long, nested sentences collapse when spoken aloud. They are hard to say, hard to follow, and easy to lose.
Short sentences create space. Each one delivers only one idea.
If I have a list of reasons or multiple parallel points, I turn each one into its own sentence. This separation serves two functions: it emphasizes each point, and it removes unnecessary cognitive load from the audience.
Sometimes logic itself needs to be explicitly surfaced. For example:
“If the meaning stays similar, consistency is high.
If the meaning drifts, consistency is low.”
Two short lines and a clear contrast.
This style helps listeners process the structure of your reasoning without straining to reconstruct it.
When drafting the script, I often—again, personally—ask ChatGPT to help shape phrasing. I feed it the ideas I want to convey and ask it to produce a speakable version. But the responsibility for the content remains mine: I revise each paragraph, imagining a skeptical listener asking why something is true, whether the logic is complete, or whether a claim is stronger than the evidence allows.
I also check for accessibility. If someone with no domain background listened to this talk, would they follow? If an expert listened, would they feel that my logic is consistent and my claims proportionate? Balancing these two audiences forces the script into its cleanest form.
⸻
4. Delivery: Pacing, Presence, and Practical Tools
By the time I reach delivery practice, the slides are structurally sound and the script is clear. Delivery then becomes an exercise in pacing, rhythm, and presence.
I follow advice that many presenters recommend, but that I have found to be genuinely impactful:
Speak slower than you think you should. Nerves speed you up. Deliberate slowness keeps your thoughts intact and gives the audience space to absorb them.
Pause at difficult points. Technical terms or subtle ideas require processing time. A short pause—one second, sometimes two—helps the audience catch up and helps me reset my pace.
Mark difficult lines in the script. During practice, I underline or highlight terms I know I tend to rush through. These marks remind me to slow down or pause.
Read the script out loud, line by line, and revise as you go. A sentence that looks fine on the page may be awkward when spoken. Speaking it aloud forces you to confront rhythm and clarity issues early.
Use deep breathing when nervous. This is more than a cliché. Deep breathing interrupts the physiological cycle that accelerates speech, tightens the throat, and scatters attention.
The goal is not to sound dramatic or polished, but to sound stable.
Stability is what makes a research talk feel trustworthy.
⸻
Closing Thoughts
Preparing a research talk is not a performance of intelligence. It is a construction project. Slides shape structure; scripts shape language; delivery shapes pacing. These layers support each other, and clarity emerges only when all three are developed with intention.
If I had to summarize everything I’ve said, it would be this:
Clarity is not something you stumble into.
It is something you build—decision by decision, layer by layer.