How to turn a paper, a lab report or a dataset into a simulation video

A simulation video only helps if the model really runs. How to turn a paper, a lab report or a table of results into one, and how the options compare.

By the Chasca team 5 min read

This video was made with the Simulation lab skill. Read the transcript

A simulation video shows a model running: an outbreak spreading through a school, hares and lynx rising and falling, a sample decaying in a lab. It only helps if the model on screen really runs, with every peak, curve and number computed from its equations and starting values rather than drawn to look plausible. To make one from a paper, a lab report or a dataset, you need the model’s equations or rules, its settings and starting values, one question the video answers, and a tool that runs the model before it narrates it.

What makes a simulation video different from an animation?

An animation is drawn. A designer, or an AI video generator, decides where the curve peaks and how fast the particles move, and the result looks scientific. A simulation is computed: the model runs from its equations, and everything on screen is read from that run.

The difference matters most when the video makes a claim. If a narrator says the epidemic peaks on day 12 with 260 pupils ill, a drawn curve can peak anywhere. A computed one peaks where the model says, so the picture, the chart and the words agree, and a viewer who checks the numbers finds them right. An AI text-to-video model can’t promise this, because it generates pixels, not solutions.

Does watching a model run help people understand it?

On average, yes, with conditions. Nico Rutten, Wouter van Joolingen and Jan van der Veen reviewed a decade of experimental studies of computer simulations in science teaching and found they can enhance traditional instruction, especially lab work, while a better-looking visualisation didn’t by itself mean better learning. Tim Höffler and Detlev Leutner’s meta-analysis of 26 studies found instructional animations beat static pictures (d = 0.37), most when the animation showed the content itself rather than decorating it.

Interactive simulations such as PhET’s work because they show what a still diagram can’t: a system changing over time, and the effect of changing one setting. A narrated video can’t be explored like PhET, but it can guide the viewer to the moments that matter, at the pace of an explanation, in their own language.

What can you turn into a simulation video?

Four kinds of material work well:

  • The equations in a paper. Most papers in epidemiology, ecology, chemistry and economics state their model as equations with parameters. The SIR model of an epidemic goes back to Kermack and McKendrick in 1927; predator and prey to Lotka and Volterra. Run as written, they become stocks that fill and drain, with a chart and a phase plane drawn as they change.
  • A lab report. A first-order reaction, a cooling cup, a half-life: the method gives the setup, and the results give the numbers the run should reproduce.
  • A table of measurements. Counts of yeast cells every two hours, temperatures over a morning, positions of a trolley. Replayed on a clock, each measurement lands on the chart as it is reached.
  • A question. Why do traffic jams appear on a road with no crash and no lights? Sugiyama and colleagues showed in 2008 that 22 cars driving round a circular track form a jam with no bottleneck at all. A ring road where every driver follows a few simple rules shows the same thing.

How do you make one, step by step?

  1. Pick one question. “Why did the flu peak on day 12?” or “What would closing the school for a long weekend have changed?” A video that answers one question is easier to follow than one that tours a whole paper.
  2. Collect the model. The variables and their starting values, the parameters with their units, the equations or rules, and the time span. With data, the table with its columns named.
  3. Choose the comparison. Most findings are comparisons: with and without an intervention, warm against cold, two flasks. Two runs side by side, on one scale, teach more than one.
  4. Write the words, or let the material lead. Your own script, narrated as written, or an explanation that follows the paper in its own order and terms.
  5. Check the numbers. Every result the paper reports should be reproduced by the run. If one isn’t, find out why before anyone else does.
  6. Keep it short. Philip Guo, Juho Kim and Rob Rubin’s study of 6.9 million edX video sessions found engagement fell after about six minutes. One to three minutes per finding is a good target.

How do the options compare?

Way to make it What it gets right What it costs you
Code it (Python, SciPy, Manim) Exact, and every detail is yours Days of work, and the voice and timing are separate jobs
Interactive simulations (PhET, NetLogo) Free, explorable, well tested A fixed set of models, or modelling skills; not a narrated video
AI text-to-video generators Fast, good-looking footage Nothing is computed, so numbers and curves won’t match a real model
Simulation lab in Chasca The model run exactly from your words or material, its chart from the same run, narrated in 79 languages A finished video to watch, not a simulation to play with

Code when students should change the settings themselves, or when your model needs something no tool offers. Use an interactive simulation in a lab lesson. When you want a finished, narrated explanation of a model or a result, make a simulation video.

Making one with Chasca

Chasca’s Simulation lab skill takes a question, your own script, or a paper, a lab report or a table of results (PDF, document, CSV or Excel). It reads the model and runs it before anything is said, then narrates while the world plays on screen and the chart grows from the same run. Results your material reports are checked against the run, and the video’s page lists what matched and what differs. It runs epidemics, markets, populations, traffic, projectiles, pendulums, springs, orbits, gases, diffusion, heat and waves, as well as the equations a paper states. A model it can’t run is refused with the reason, never swapped for a different one. The free videos are up to a minute long, so you can try one with your own paper.

Video transcript What the narrator says in the 1:22 explainer above.

The study follows flu through 840 boarding school pupils, beginning with two ill pupils.

The susceptible group can still catch flu. Infection moves pupils into the ill group as they meet.

Recovery moves ill pupils into the recovered group, who no longer catch flu in these equations.

The infectious period averages 2.4 days. The contact parameter controls how quickly susceptible pupils become ill.

The morning bed counts appear as points, letting you compare the study’s counts with the model’s curve.

Without closure, illness peaks at 361 pupils around day 5.3, then falls as susceptible pupils become scarce.

By day eighteen, 828 pupils have had flu, including recovered pupils, not just those still ill.

For the closure scenario, the study lowers the contact parameter from 1.8 to 1.1.

This scenario keeps contacts reduced throughout the run, rather than restoring them after the weekend.

Illness now peaks at 214 pupils on day 9.6, lower and later than without closure.

The comparison therefore shows sustained contact reduction, not specifically what closing for a long weekend would achieve.

Questions people ask

Yes, if the tool runs the paper's model rather than drawing pictures of it. Chasca's Simulation lab reads the equations or rules, parameters and starting values from a paper, runs them, and narrates the video in the paper's own order while the model plays on screen, with a chart drawn from the same run. Results the paper reports are compared with the run, and the video's page lists what matched.

An animation is drawn to look right: a curve rises and falls because someone drew it that way. A simulation video is computed: the model runs from its equations and starting values, and every position, peak and number on screen comes from that run, so the picture, the chart and the narration always agree.

The evidence says they can. A review of a decade of experimental studies found computer simulations can enhance traditional science teaching, especially lab work, and a meta-analysis of 26 studies found animations beat static pictures when they show the content itself. How the simulation is guided matters as much as how it looks.

Ordinary differential and difference equations such as decay, logistic growth, SIR epidemics and Lotka–Volterra predator and prey; agent worlds such as epidemics, markets, natural selection and traffic; physics worlds such as projectiles, pendulums, springs, orbits, gases, diffusion, heat and waves; and tables of measurements, replayed point by point.

Usually one to three minutes: long enough to set up the model, run it and show one comparison. A study of 6.9 million video sessions on edX found engagement fell after about six minutes, so a longer paper is better split into one video per finding.

Keep reading

Your next explainer takes under a minute.

Your first 2 videos are free, up to 1 minute each, watermarked. No card required. Full product, every voice and style included.