---
name: "convention-visual-explanation"
type: convention
description: 'Which visual format explains a given idea, and the quality bar each format is held to. Covers four formats — diagram, animated diagram, conceptual illustration and moving image (explainer and promo cuts) — with the format-choice decision, the bar per format, the bars common to all motion, the sound bar, and the rule that every figure on screen traces to a source. Read before commissioning or producing any explanatory visual, so the bar arrives with the brief instead of being rediscovered per job. Cross-venture by intent: the bars here are venture-neutral and the look is not — each venture''s palette, type and art direction live in its instance in convention-brand, and the production sequence lives in skill-visual-production. Owned by Pixel.'
license: CC BY-NC-SA 4.0
---

# convention-visual-explanation

Nothing in canon said which visual format a given idea needs, or how good each one has to be. So the bar was rediscovered per job: the Mr Pritchard visual session of 2026-10-21/22 spent four rounds on hero concepts, rewrote a storyboard in full, and invented a second video format before the work met a bar nobody had written down. This file writes it down. (APP-2596; Owner, 2026-10-22: "proposal yes. go ahead".)

**What this file is not.** It carries no palette, no typeface and no art direction. The bars here hold for any venture; the look does not. The look lives in the venture's instance in `convention-brand`, and the procedure that produces the work lives in `skill-visual-production`. Read all three for any explanatory piece: this one for the format and the bar, the brand instance for the look, the production skill for the sequence.

**Inheritance.** This sits beneath `convention-aesthetic`, the cross-venture taste layer, like every other visual artefact — define against it by divergence rather than restating it.

## Choosing the format

Choose on what the idea needs shown, never on what is cheapest or most impressive to make. Four formats, and the question each one answers:

- **Diagram** — *the idea has a structure, and the reader needs to see the parts and their relations.* Static, read in an article body at the reader's own pace. This is the default: reach for it first, and move up only when something below is genuinely needed.
- **Animated diagram** — *the structure has an order, and the order is the point.* Choose this over a static diagram when the argument is built rather than merely laid out: a sequence, a cause, a thing that happens in steps. 10 to 15 seconds. It must end on, and hold, the static diagram, so the still stays usable and a reader who misses the build loses nothing.
- **Illustration (hero)** — *the piece needs a way in, not an explanation.* A conceptual metaphor at the top of an article or post. It is **not** the diagram in another style, and not the diagram drawn in isometric form (APP-2554, taste signal; Owner, 2026-10-21: "the hero image is not linked to the diagram but is an illustrative representation").
- **Moving image** — *the idea is a narrative, or it needs an audience it does not yet have.* Two cuts, chosen on the job:
  - **Explainer** — carries the argument. Narrative order: situation, risk, solution, benefit, outcome (Owner, 2026-10-21).
  - **Promo** — earns attention for the explainer. Picture-led, almost no text.

Where a diagram would do, a film is not an upgrade: it is a slower, costlier diagram that most of the audience will not finish.

## The bar per format

### Diagram

Plain-English labels — no internal vocabulary a reader outside the work would have to decode. One grid, held across the whole set. **One meaning device**, used consistently: whatever marks the human, the decision or the focus does that one job and no other. Legible on a phone.

### Animated diagram

Everything above, plus: the build order *is* the argument's order, step for step. One thing enters at a time. It ends on the static diagram and holds it long enough to be read. 10 to 15 seconds — a build that needs longer is an explainer in the wrong format.

### Illustration (hero)

A conceptual metaphor, not a depiction of the mechanism. **Vary the metaphor type across a series** — object, machine, figure — rather than running one idea through every piece (Owner, 2026-10-21: "why do they all follow the same idea? brainstorm conceptual ideas"). **Offer three concepts per piece before anything is drawn**, and let the Owner pick: three variants of one metaphor type is one concept, not three. No text set inside the illustration.

### Moving image — explainer

Every line holds on screen **at least 3.5 seconds**. One continuous world: move the camera rather than cutting between unrelated spaces. Type sits on clean paper and is never crossed by a scene line. Readable on a phone, which is the only screen size that settles whether type is big enough.

### Moving image — promo

Picture-led. Type appears only on a title card, an interstitial, a short overlay or a sign-off card. **At most five words on screen** at once. **Card type at least 8% of frame height.** The promo cut exists because the explainer failed this test: the Owner's verdict on it, 2026-10-22, was that it was "difficult to read due to the amount of text". A promo that reads like a short explainer has not been made.

## Bars for all motion

Owner, 2026-10-21: *"The key to the visual / motion is that it helps explain the concept better than a static image and engages the audience more than words."* That is the test motion has to pass, and a handsome film can fail it. From it:

- **Motion shows cause and effect** — something moves because something else did. Movement that is only decoration fails the test.
- **One focal point at a time.** Two things competing for attention is nothing moving.
- **Type on clean paper, never crossed by scene lines**, and sized for a phone.
- **A closing image must belong to the argument.** An image that is beautiful but bolted on is cut, not kept because it is good. (Owner, 2026-10-22: the loom ending was cut as "bolted on" and kept as a standalone illustration.)
- **Where the work draws an assistant, it is drawn as the wrapper.** The assistant surrounds the agents and the operators and manages the one door through which work reaches the Owner, who stays outside it. A drawing that puts the Owner inside the system, or the assistant alongside the agents rather than around them, is wrong about the thing it is explaining. (Owner, 2026-10-22.)

## Sound

**One continuous score that follows the picture.** The score tracks what is on screen and tightens or eases as the picture does. It does not change register.

**An abrupt change of musical register mid-film is a named failure.** This bar was tried the other way and rejected. A two-part score was built for the Mr Pritchard promo — darker under the problem set-up, moving to a calmer then upbeat register at the turn to the solution — and on hearing it the Owner ruled, 2026-10-22: *"music is awful when it cuts to that happier version so let's keep it with the original music."* The original single-register score, following the picture throughout, is the one he kept. So do not score a film in two halves, and do not read "the narrative turns here" as a licence to switch register: the turn is carried by the picture and by rhythm density, not by a different kind of music.

Also:

- **Every sound is caused by something on screen.** No sound without a source in the picture.
- **Tension comes from rhythm and density**, not from volume or from a new instrument arriving.
- **No dead silences.** A gap reads as a fault rather than as a pause.
- **Sound is approved by ear, never by measurement.** A score analysed in code is unverified until the Owner has listened to it, and it is reported that way.

## Kinds of on-screen text

Text in a motion piece failed in both directions on the same film before this was written: the explainer carried too much of it, and the promo, stripped back to cards, lost the meaning the text had been carrying. Both are the same defect — text doing a job nobody had assigned it. Three kinds, and each answers exactly one question.

**The line** says *why this scene matters* — the argument. One full sentence per scene, at most fifteen words, set top left on clean paper. Its bars are the explainer's, above: at least 3.5 seconds on screen, never crossed by a scene line.

**The figure** says *how much* — a number that changes while the viewer watches. Large numerals with a short label attached to the object, no box, counting as the scene builds. Use one only where the change itself is the point; a number that merely sits there is a label. Every figure traces to a source, per *Evidence* below.

**The annotation** says *what that is* — and only where the picture alone is ambiguous. A paper box with an ink hairline border and a leader line to a dot that tracks its object in three dimensions. At most eight words. One at a time, two at the very most. Held for (words ÷ 3) + 1 seconds, and the picture slows while it is up, because an annotation the viewer has to read against a moving frame is an annotation they skip. The paper box is the form; the inverted box was tried and rejected, because it pulled attention away from the thing it was pointing at. Type size and face come from the venture's instance in `convention-brand`, never from here.

The hierarchy, which is what stops the three competing:

1. **One question each** — the line *why*, the figure *how much*, the annotation *what is that*. A kind answering a second question has become a different kind badly.
2. **At most two kinds on screen at once**, read in order: line first, then the figure or the annotation.
3. **Never say the same thing twice across kinds.** If the line says it, do not annotate it.
4. **Annotate only where a viewer would ask "what is that?"** — not wherever a thing has a name.

By format: an **explainer** uses all three. A **promo** uses cards, figures and annotations and carries no full lines, which is the whole of the difference between the two cuts. An **animated diagram** uses figures and labels only. A **static diagram** uses labels only.

## Evidence

**Every figure on screen traces to a ticket or a truth capture.** This is `convention-mrp-voice`'s evidence rule, which governs what a piece says, extended to what it shows: a number on a card is as much a claim as a number in a sentence, and a frame is the easier place to lose the source. A figure with no trace comes off the frame.

## Guardrails

- Choose the format on what the idea needs shown, never on what is most impressive to produce. A film where a diagram would do is a downgrade dressed as an upgrade.
- Never let a hero illustration become the diagram in another style, and never run one metaphor type through a whole series.
- Never draw before the Owner has picked from three concepts, and never render a film before he has approved the storyboard.
- Never score a film in two registers. One continuous score, following the picture.
- Never report sound as approved on measurement alone.
- Never put a figure on screen that does not trace to a source.
- This file holds bars, not looks. A render needing a palette, a face or an art direction takes them from the venture's instance in `convention-brand`, and stops and says so where that instance has a gap there — the tool default is never the answer (`convention-aesthetic`).
