A vertical video of 30–60 seconds that answers one question, like "Why does ice float on water?" Narration and captions carry it, and the pictures show what is being said.
In Shot 1, fill the frame with ice floating in water and put the caption 'Why does ice float on water?' on the first frame. The question and its subject appear together, so viewers know right away what will be explained.
To recreate it: Include only the subject of the question in the hook shot and no other objects.
In Shots 1 and 5, show the ice very close, down to the trapped bubbles and splashing droplets. Even a familiar object becomes intriguing at this distance.
To recreate it: Set a background color like 'dark blue background' and the subject stands out clearly.
In Shot 2, show ice crystals spreading across the water surface at large scale. This shot turns the moment the narration says 'when water freezes' straight into a picture.
To recreate it: For slow changes like spreading crystals, about 3s on a fixed camera is enough.
In Shot 3, place the two glasses in the lower half of the frame and leave the top empty as room for captions. The explanation reads even with the sound off.
To recreate it: AI draws garbled text in the picture, so write 'no text, empty space in the upper half' and add the text in editing.
Lay the narration first and place a matching shot over each sentence. Words and pictures are paired, so it is easier to understand.
To recreate it: Write the script first and generate one shot per sentence separately.
Snap the picture in to 110-120% when a key word comes up. It works as a highlight, and the same shot doesn't get boring even when it runs long.
To recreate it: Zooming in lowers image quality, so generate the shots you'll punch in on at a high resolution.
End Shot 5 with the same background and the same ice size as Shot 1, so the last scene connects straight into the first. With no sense of a break when it replays, people watch it again and again.
To recreate it: Make Shot 1 and Shot 5 from the same start-frame image to match the background and light.
Cinematic multi-shot sequence, 13 seconds total, vertical 9:16. Shot 1 (2s): extreme close-up slow push-in on a clear ice cube floating in water, tiny air bubbles inside; Shot 2 (3s): static macro of ice crystals spreading across still water, cool blue backlight; Shot 3 (3s): static shot of two clear glasses in the lower half of a white table, large empty space above for captions; Shot 4 (3s): aerial drone shot gliding over a frozen lake, wide framing, dark water under thin ice; Shot 5 (2s): static slow-motion extreme close-up of an ice cube dropping into a glass, droplets frozen in the air. Consistent clean look, bright high-key light, cool blue palette, crisp detail, no text, curious mood.
Fill the first frame with the subject of the question, and people know right away what is being explained. Pair each sentence with one picture and leave room for text, and it still makes sense with the sound off.
Vertical 9:16 extreme close-up of a clear ice cube floating in a glass of water, slow push-in, tiny air bubbles trapped inside the ice, bright clean side light against a dark blue background, crisp detail, curious intriguing mood.
Vertical 9:16 macro shot of ice crystals slowly spreading across the surface of still water, static camera, cool blue backlight revealing branching patterns, extreme detail, shallow depth of field, clean scientific look, calm fascinating mood.
Vertical 9:16 static shot of two clear glasses side by side on a white table, ice floating at the top of one, glasses placed in the lower half, large clean empty space in the upper half for captions, bright high-key lighting, minimal clean look, no text.
Vertical 9:16 aerial drone shot slowly gliding over a frozen lake, dark water visible under thin ice near the shore, wide framing, soft overcast light, cool desaturated blue tones, crisp detail, quiet wondrous mood.
Vertical 9:16 extreme close-up of an ice cube dropping into a glass of water in slow motion, droplets frozen in the air, static camera, bright clean side light against a dark blue background, crisp detail, satisfying mood.
Put the question up as a hook caption in the first 1–3 seconds, from the very first frame. Lay the narration down first, then add one B-roll overlay cut for each sentence. When a key word is spoken, pull in with a punch-in of only 110–120%, and keep caption styling to two lines or fewer inside the safe area. End with a seamless loop, so the last shot connects to the first and people watch again.
Write the script first, generate one shot per sentence separately, and write only one camera move per shot. AI draws text, numbers, and charts as gibberish, so leave space for text in the start frame and add the explanatory text in editing. Scenes that contradict the facts, like ice sinking, are easy to get, so first check in a 480p draft that the content is correct.