Animated Teaching Visuals (B-Roll) for Lecture Videos

The finished effect

What it looks like when it is done.

Twenty-two seconds of the 22-minute Indian Wisdom lecture used throughout this page (6:00 to 6:22 of the lecture): three drawn cutaways in a row — a versus and two focus scenes — cut into Presenter + Animated, each on the phrase it illustrates, with the caption band carried straight over them. Every title has at least 88 px of clear space under it before the drawing begins.

22 s · 1920 × 1080 · with sound

Overview

What it is, and why it exists.

An Animated Teaching Visual is a full-screen cutaway, three to seven seconds long, that leaves the lecture's current view — the full-frame deck, the presenter beside the slide, or the presenter with captions — for an original illustration of what is being said, then returns. It is one of three kinds of cutaway, alongside generated stills and generated video clips, and it has its own budget so it never competes with them for the same seconds.

It exists because most of a lecture cannot honestly be photographed. A roadmap ('by the end of these twenty minutes'), a definition, a distinction between two schools of thought: asking an image model for any of these is how a lecture on human resources ends up illustrated with a field of cows. A drawn scene can carry the idea itself, it costs nothing to make, and it cannot fabricate a person, a place or a document.

The composition plan decides where each drawn cutaway goes and which motion treatment it gets; a short scene description for the lecture decides what is drawn there — a shape, a kicker and headline, icons and short lines, each tied to the phrase that introduces it; and a dedicated renderer draws each one to its own clip, which is dropped into the film under a fresh caption band.

It depends on the word timeline, the composition plan's phrase boundaries and budgets, the course's template style, and the loudness envelope. Every cutaway is named by its own start time in hundredths of a second — the cutaway at 58.19 s is 5819 — so the plan, the scene description, the renderer and the assembler can never disagree about which clip is which.

1 a minute, its own quotaBudget

14% of the runtimeScreen-time ceiling

2.8-7.5 s, ideal 5 sClip length

4, never the same twice runningTreatments

row, versus, focus, chainShapes

17 plain icons, no letteringIcon set

one move per clip, up to 6% pushCamera

none — drawn, not generatedCost

How it is made

Step by step, in the order the engine runs it.

Find the candidate windows

The planner lists every window of one, two or three consecutive caption phrases between 2.8 s and 7.5 s long, starting after the first 15 s and ending before the last 25 s, not straddling a beat or layout edge and at least 1 s clear of any layout change. Each is scored for concreteness, lists, comparisons, processes and numbers, and penalised for pure connective phrases.

Decide what may be generated and what must be drawn

A window that names a real entity or reads as a document, chart, statistic or headline can never be a generated picture; it can only be drawn. A window with nothing a camera could point at is also drawn. Everything else is eligible for a generated still or clip.

Place drawn cutaways against their own quota

Stills and drawn cutaways take turns, each serving whichever kind is furthest behind its per-minute target, so neither eats the other's schedule. Drawn cutaways are held to 14% of the runtime. No cutaway may overlap another, start within 6 s of another's start, crowd more than 14 into any 120 s, or land on a definition, quote or hero.

Close the gaps

A final pass walks every stretch with no cutaway. The first picture must arrive by 90 s and no stretch may run 60 s without one. Where the ordinary pool has nothing for a stretch, a wider pool with no score bar is used — and anything taken from it is forced to a drawn treatment, never a generated photograph.

Choose a treatment

Each drawn cutaway is given one of four treatments by rule: Progressive Build-On if the phrase lists, sequences or compares; Pan & Zoom if it is under 6 s; Floating & Pulsing Glow if it sits over the caption layout; otherwise Parallax, then the others in turn. A treatment is never used twice in a row.

Compose the scene

For each drawn cutaway the scene description names a shape (row, versus, focus or chain), a camera move (push, pull or drift), a kicker, a headline and the items — each with an icon and the phrase that introduces it. Every phrase is looked up in the transcript starting 4.5 s before the clip, converted to seconds from the clip's start, and a phrase that cannot be found stops the build.

Draw each clip

The renderer draws each scene at 30 fps for exactly the cutaway's length, in the course's style and in the light or dark palette of the chapter it lands in, and writes one clip per cutaway id. A clip that already exists and is complete is kept; one smaller than 10 KB is redrawn.

Cut it in, with captions on top

The assembler takes the clip for each cutaway id, fits it to the 1920 x 1080 frame, and overlays the caption band drawn fresh for that window. The narration underneath is untouched. If a clip is missing, the piece falls back to the layout underneath and the assembly record says so.

Elements

Everything this effect puts on screen, one piece at a time.

Each element below is shown as the engine actually produced it, in the real lecture, with its parts, and the rules that govern it.

Pan & Zoom Progressive Build-On Parallax Layers Floating & Pulsing Glow Row Versus Focus Chain Kicker and headline Icon card Icon vocabulary Consequence pill Ground Camera move Dual-tone palette Caption band over the cutaway

0:30 — 'what this course gives you': a slow push on the mortarboard.

Pan & Zoom

Treatment: the camera eases toward the payoff

Pan & Zoom treats the illustration as an oversized still and moves a crop across it, easing in toward the element the phrase is building to — a culminating idea, a checkmark, the last item in a short sequence. The move follows a smoothstep curve, so it starts and ends at rest and never jolts at the cut. It is the planner's first choice for any cutaway shorter than 6 s. In the drawn-scene renderer the same idea is carried by the push: the whole scene scales from 100% to 106% over the clip. In the lecture below it opens the drawn cutaways at 0:30, pushing in on a mortarboard as the lecturer says what the course will cover.

Anatomy

Reference treatment: illustration built at 2304 x 1296 (1.2x the frame); crop eases from the whole canvas to a 1340 x 754 box on the payoff
Drawn-scene push: scale 1.00 to 1.06 around the frame centre, smoothstep over the whole clip
Pull variant: 1.06 back to 1.00