Most physics apparatus is abstract. A Van de Graaff generator is not. It makes hair rise, it throws a snapping spark across a gap, and it turns an invisible force into something a room full of people can hear and see. That combination of visual payoff and genuine scientific depth makes it one of the best subjects an educator can put on camera.
Video also solves problems the live demonstration cannot. A live demo happens once, in one room, at one moment, and the students in the back row see almost nothing. A recorded demo can be repeated, slowed down, zoomed in, annotated, captioned, and paused for discussion. It can be watched in a classroom with a substitute teacher, at home on a phone, or in a museum gallery on a loop. A good recording multiplies the reach of a single afternoon of setup.
But a recording is not automatically a lesson. Pointing a camera at a working generator produces footage of a working generator. Turning that footage into teaching material requires decisions about what the viewer needs to see, what they need to hear, what order they need it in, and where the camera must go to make the invisible visible. This guide walks through the full pipeline: planning, scripting, lighting, safety, capture, editing, and the places where AI tools genuinely speed things up.
Before planning a single shot, be clear about the physics you are trying to make visible. A Van de Graaff generator uses a motor-driven belt to carry electric charge up to a large hollow metal dome. Charge accumulates on the outside of the dome, raising the potential difference to a very high voltage. When the field becomes strong enough, the air between the dome and a nearby grounded object breaks down, and the charge discharges as a spark.
The problem is that charge itself is invisible. Voltage is invisible. Current is invisible. Viewers can only reason about these concepts through visible proxies, and the entire visual plan of your video should be built around supplying those proxies.
Useful visible proxies include:
The second thing viewers need is scale and context. A spark looks impressive only if the viewer knows how large the gap is and how large the dome is. A wide shot with a person standing beside the machine establishes size; a close-up sparks curiosity. Keep track of both.
Finally, viewers need the voltage-versus-current distinction, because it is the reason the demonstration is safe at all. A machine like this can reach hundreds of thousands of volts while delivering only a tiny current, and it is the low current that keeps the shock non-lethal in normal use. That nuance is the single most valuable thirty seconds of teaching in the whole video, and it is frequently skipped.
A demonstration video earns attention by showing something first and explaining it second. A structure that works well in classrooms runs roughly like this:
Keep the main version between four and eight minutes for classroom use. Then cut a 45-60 second teaser and a vertical version using only the payoff and one explanation beat.
A storyboard here is not an art exercise; it is a map between concepts and shots. Sketch a rough frame for each shot and write the concept it carries underneath it:
Also board the shots you will not be able to film: electron flow animations, field line diagrams, and labeled cross-sections. Planning them in advance tells you which frames need clean, uncluttered backgrounds so graphics can be laid over them later. That single decision saves hours of rotoscoping.
Sparks and reflections are the two things that ruin footage of electrostatic demos. A polished dome is essentially a mirror, and a softbox aimed at it will produce a blown-out white blob that reads as an error rather than a machine. Handle it deliberately:
For camera work, a locked-down tripod is your default. Sparks last milliseconds and happen unpredictably, so framing must already be correct. Use manual exposure and manual focus with focus peaking enabled, and turn off any auto-gain that will brighten the frame when the spark flashes.
A practical shot list for one session:
Audio deserves the same planning. Belt squeak and motor hum are loud and constant. Capture a clean wild track of the room, place a lavalier on the presenter, and consider recording narration separately in a quiet space. The moment of discharge is worth isolating: capture it cleanly if you can, and plan to use it as a punctuation beat in the edit.
Film crews change the risk profile of any demonstration, because more people, more cables, and more distractions sit around the machine. Establish roles before anyone touches equipment: one person operates the generator, one person operates the camera, and nobody does both.
Core practices:
Dry air improves the effect and also increases the chance of small static shocks for the crew. If the room is very dry, remind everyone to discharge themselves on a grounded surface before touching gear.
Sparks. A spark is brief and bright. To record it well, use a fast shutter speed in the 1/500 to 1/2000 range and record at a high frame rate, then conform to a normal playback rate in the edit. High frame rates demand a lot of light, so plan strong, flicker-free continuous lighting rather than flash. Shoot the discharge many times; consistency comes from repetition, and small changes in gap distance, humidity, and dome charge change the result. A dark, simple background is non-negotiable.
Hair. Hair rise reads best from a slightly low medium shot with a rim light. Ask the participant to stand still and breathe normally, count down aloud, and hold the pose for several seconds after full charge so the editor has room. A shallow depth of field separates the hair from the background, but keep the whole head in focus if you plan to overlay labels. Capture at a moderate frame rate to keep motion smooth without needing enormous light output.
Effect visibility. Humidity is the enemy of every electrostatic demo. Film on a dry day, or in a room where the ventilation has been off long enough for the air to settle. Avoid running forced-air heating or cooling during takes, since airflow both changes humidity and moves hair that you want to move for physics reasons only.
Inserts and cutaways. Build a small bank of supporting shots so the edit never has to repeat footage. The electroscope leaf, the water stream, the candle, and the glowing tube are all fast to shoot and give the editor something to cut to while narration explains.
Logging. For every take, note the frame rate, shutter angle, gap distance, and whether the shot is a keeper. On a two-hour shoot you will accumulate dozens of clips that look nearly identical. A simple paper log or a spreadsheet makes the edit dramatically faster.
Pacing is the difference between a video that teaches and a video that performs. Resist the urge to cut on every beat. Let the demo shots breathe at real speed so viewers understand cause and effect, then use slow motion only for the moment worth isolating: the instant of breakdown, or the first second of hair lifting.
A comprehension-first edit usually includes:
Mix audio so the machine hum sits under narration rather than on top of it. Duck the hum during explanation and let it rise during un-narrated demo shots. If the spark was recorded cleanly, place it exactly on the cut so the visual and audio land together.
AI tools have changed the economics of educational video, but the useful ones are the boring ones. The biggest time savers are transcription-based workflows: automatic speech-to-text gives you a searchable script, and text-based editing lets you cut the video by deleting sentences rather than scrubbing a timeline. For a talking-head lesson, this alone can halve the edit time.
Other genuinely practical uses:
Where AI hurts: interpolated frames can invent artifacts around a spark, making it look like a flicker rather than a discharge. Generated imagery can imply a phenomenon that was not actually recorded, which is a serious problem in science communication. Tempting as it is, never let a synthetic spark stand in for a real one without a clear label. Keep your original footage, verify every caption, and treat generated diagrams as illustrations rather than evidence.
A reliable AI-assisted workflow looks like this: transcribe, rough cut by text, stabilize and denoise the keepers, mask the elements that need graphics, build diagrams, match color across takes, generate and correct captions, then export a set of variants. Each step takes minutes rather than hours, and the saved time goes into pedagogy, which is where the real value lives.
A single well-planned shoot day can supply an entire term of content if you plan the exports up front. From one session you can produce:
Decide on these deliverables before the shoot, because each one changes what you need in the can. Vertical cuts require headroom and a framing that survives a crop. Stills require a shot with enough resolution and sharpness to survive enlargement. Prediction clips require the presenter to address the camera directly at least once.
Mistakes that repeatedly show up in physics demo videos:
Frequently asked questions
Is a demonstration like this safe to film with a full crew? With the standard precautions, yes. The key risks are distraction and carelessness, not the machine itself. Assign one operator, brief everyone, ground the dome before handling, and keep grounded metal away from the charged person.
Do I need an expensive high-speed camera? No. A camera that records 120 or 240 frames per second is enough for a convincing spark, and many phones now offer that. If you only have 60 fps, shoot the discharge repeatedly and use a clean take at normal speed; the cut still lands.
Can a phone handle the whole production? For a classroom clip, often yes, if you lock exposure manually, use a tripod, and record audio separately. The limits show up in low-light slow motion and in audio quality.
Where should the machine sit in the frame? On a dark, uncluttered background with the dome in the upper third and the ground wand entering from the side. Leave space opposite the spark for labels or a scale marker.
How do I explain voltage versus current without confusing viewers? Use a water analogy with a narrow pipe: high pressure, low flow. Show the voltage number on screen, then show the tiny current number beside it, and say plainly that the second number is why the demonstration is safe.
Should I use real footage or animation for the invisible parts? Both. Real footage establishes that the phenomenon happens; animation explains why. Label animation clearly and never present it as captured footage.
What if the spark refuses to appear on camera day? Check humidity, gap distance, and dome cleanliness first. A humid room, a gap that is too wide, and a dusty dome are the three most common culprits. Keep a small, controlled indoor gap as a backup setup so you always leave with at least one usable discharge.
Do I need consent to film participants? Yes, and especially for minors. Collect written permission, explain what will happen, and give people an easy way to decline without disrupting the shoot.
A Van de Graaff demonstration rewards planning more than almost any other piece of science equipment, because almost everything interesting about it is either invisible, instantaneous, or both. Build the script around visible proxies, light it so the spark has somewhere to appear, shoot the discharge more times than you think you need, and edit for comprehension rather than spectacle. Do that, and one afternoon with a generator becomes a lesson that keeps working for years.

Alexander
Vidu Q4 alternatives compared: Kling 4.0, Seedance 2.5, Wan 3.0, and Runway Gen-4 — 4K support, input modes, durations, audio, and what each costs in credits on Domer.
The top 4K AI image generators on Domer compared: Flux 3 Klein, Nano Banana 2.1, GPT Image 2.5, GPT Image 2, and Nano Banana Pro — 4K credit costs, reference limits, and speed.
Nano Banana 2.1 vs GPT Image 2.5 Flare compared: credit costs per resolution, aspect ratios, reference images, text rendering, and which tasks fit each model.