To illustrate an abstract idea, do not try to picture the idea. Picture the shape of the claim you are making about it. Almost every abstract claim is structurally one of a small number of things: a contrast, a sequence, a containment, a proportion, a loop, a threshold, or a network. Each of those has an obvious drawing, and the drawing teaches, because it shows the structure the words are describing. There is no picture of alienation, no photograph of sovereignty, no stock clip of moral hazard. There is a very good diagram of every argument you are likely to make about them.

This is the hardest part of making explainers about ideas, and it is where most of them fail. The failure is not ugly visuals. It is visuals that are merely present.

The three bad defaults

Before the method, the three things almost everyone reaches for first, and why each one fails.

The literal metaphor. A lightbulb for an idea, a chess piece for strategy, a maze for complexity, scales for justice. These are not wrong so much as empty: they illustrate the category of thing you are discussing rather than anything you are saying about it. A viewer learns nothing from a lightbulb they did not already know from the word "idea."

The mood shot. Footage of a city at dusk under a paragraph about late capitalism. It sets a tone and carries zero information, and if the footage is generated rather than stock, it carries zero information at higher resolution. A cinematic shot of "data flowing through a network" cannot hold your labels, your sequence, or your numbers, because it was imagined around your content rather than constructed from it.

The sentence on screen. Putting the narration up as text feels safe and is actively harmful. When a viewer reads the same words they are hearing, the two channels compete rather than reinforce, and comprehension drops. This is the redundancy effect, and it is one of the better-established findings in multimedia learning. Spoken words plus a picture beats spoken words plus the same words written down.

What all three have in common: the visual is decoration next to the argument rather than a second statement of it.

Draw the shape of the claim

Here is the move. Take the sentence you are about to narrate and ask what kind of statement it is. Not what it is about, what it does. There are not many answers.

Nine shapes covers a startling amount of philosophy, geopolitics and economics. The reason is not that ideas are simple. It is that arguments have structure even when their subjects do not, and the structure is what the viewer needs to hold.

Worked examples

Deleuze and Guattari, the rhizome. People say there is no diagram for Guattari. In fact A Thousand Plateaus opens with one of the most drawable claims in twentieth-century philosophy: arborescent structure against rhizomatic structure. The tree has a root, a trunk, a single path between any two points, and a hierarchy of branches. The rhizome has no center, any point connects to any other, and cutting it anywhere leaves it able to regrow. That is a contrast, and the drawing is a tree next to a tangle. Draw them side by side, label the properties as the narration names them, and you have taught the distinction better than the paragraph does, because the paragraph has to describe in sequence what the picture shows at once.

The security dilemma. A state arms itself for defense. Its neighbor cannot tell defense from preparation, so it arms too. The first state reads that as confirmation. This is a loop, and it should be drawn as one: four nodes in a circle, arrows completing as the narration completes them, and the thing the viewer takes away is that no one in the loop is required to be hostile for it to tighten.

Moral hazard. Someone makes a decision while someone else bears its downside. That is not a loop or a contrast. It is a transfer: draw the risk as a weight, draw two parties, and move the weight. The insight lands the moment the weight lands on the wrong figure.

Comparative advantage. A proportion claim, and one that is routinely illustrated with ships and globes, which teaches nothing. The actual content is that two parties both gain by specializing even when one is better at everything. Two bars, honest relative sizes, drawn before and after. The viewer sees both totals rise.

Alienation. This one is the honest hard case, and it is worth being straight about it. There is no structural diagram of alienation that is not a lie about the concept. So do the other thing: illustrate a particular instance rather than the abstraction. One worker, one line, one object leaving their hands. The particular carries the general, which is how the concept was taught to most people in the first place.

Three questions that find the shape

When a sentence resists, ask these in order. One of them almost always produces a drawing.

1. 1

   What would have to be true for this to be false?

   This surfaces the contrast. If the claim has a real opposite, you have two things to draw, and the difference between them is the lesson.

2. 2

   What happens next?

   This surfaces the sequence. If the claim is the first half of a mechanism, the drawing is the mechanism running, and the narration can drive it one step at a time.

3. 3

   Compared to what, and how much?

   This surfaces the proportion. Abstractions often hide a quantity: 'most', 'rarely', 'the overwhelming majority'. Once a magnitude is in the sentence, there is a bar to draw.

If all three come back empty, the sentence is probably not a claim. It is a transition, or a restatement, and it does not need its own visual at all. Long-form explainers get much better the moment their maker accepts that not every sentence earns a picture.

Build the drawing in the order the argument builds

A finished diagram that appears all at once is a slide. The same diagram, drawn part by part as the narration names each part, is teaching. The difference is the ordering: the viewer's attention lands on the element being talked about, at the moment it is talked about, and the structure accumulates in the same sequence as the reasoning.

This is why hand-drawn and whiteboard styles work so well for conceptual material, and why they are not merely a nostalgic aesthetic. Drawing takes time, and the time is the point. It is the same reason a lecturer who builds a diagram on a board is easier to follow than one who projects the finished version.

How this gets made without an illustrator

The method above is the hard part and it is not automatable, because it is a judgment about your own argument. The production, however, is.

In openCanviz you paste the text and get a drafted video back: the argument split into scenes, narration written for each beat, and a constructed drawing built on screen as the narration plays, in whiteboard, doodle, painted, cutout, or poster styles. Because each scene is a structured document rather than a rendered clip, you can reach into it: move the element, relabel it, change the order in which the parts appear, and redraw the beat so that the picture matches the claim you actually meant. That reach is the whole difference between an editable explainer and a generated one, and it is explained further in what a scene actually is.

The practical loop is: write the claim, ask which of the nine shapes it is, put that shape in the text as a plain sentence ("this is a loop with four stages"), and let the draft build it. Then fix the two or three scenes where your idea had a shape the words did not reveal.

Common questions

Is it wrong to ever use a metaphor? No. A metaphor is fine when it is the argument, not when it is a substitute for one. If you are claiming that institutions ossify, a drawing of something ossifying is the claim. A lightbulb next to the word "idea" is not.

What about ideas with genuinely contested structure? Draw the contest. Two competing structures side by side, attributed, is more honest and more interesting than one confident diagram, and it is often the actual state of the field.

How much text should be on screen? Labels, numbers, and proper nouns. Not sentences. A label names the thing the narration is discussing so the viewer can bind the word to the shape; a sentence competes with the narration for the same channel.

Does this apply to short videos too? The method does; the budget does not. A ninety-second clip has room for one shape. This is really a long-form technique, which is why it pairs with turning a long essay into a narrated video.

Why not just use AI-generated footage now that it looks good? Because quality of image is not the constraint. Generated footage is imagined around your topic; a diagram is constructed from your claim. Only one of them can be relied on to put your exact labels, your exact sequence, and your exact numbers on screen.