# On Air Solid ink tags laid on someone's own footage. The person and their shots are the picture; a tag names, counts or points at what they are saying, then gets out of the way. One signal colour marks what matters. ## Essence, and what it is not - The footage is the ground; a tag sits on a shot the way a label sits on a box. - Tags are solid and dark, so they read on any shot, and small: never more of the frame than the person's head. - One signal colour, a clear teal, on the one thing a tag is about: an edge, a unit, a ring, an underline. - Things point at the footage itself: a ring round the sticker on the laptop in their hands, not a drawing of a laptop. Not a news ticker (no scrolling bars or permanent frames), not a meme edit (no stickers, emoji or shaking text), not an explainer (no scenes drawn over the person; the inset frames them and keeps them playing). ## Ground and material The recording is the ground; the motion draws none. A tag is ink (`panel` at 90 %), 4 px radius, a short drop shadow. Where a shot is too bright behind a tag, `scrim.tsx` shades that edge only; where it is too busy for any tag, `inset.tsx` frames it and lays ink beside it. Nothing is glassy, blurred or glowing. ## Colour roles - `data`: the signal teal, the one accent: an edge, a unit, a ring, an underline, the item being said, a winner, a plus. - `text`: warm white for what a tag says. `muted`: what is no longer the point (a struck word, the loser, a minus). - `bad`: only the stroke through a word that is wrong. `panel`: the ink of every tag; teal is a line or a word, never a tag's fill. ## Type and captions Atkinson Hyperlegible Next for all a viewer reads: names 50 px bold, figures 112 px extra bold with tabular digits, headlines 58 px, labels 38 to 48 px. JetBrains Mono, spaced and in capitals, for the line over a tag (26 px, teal). Chinese: Noto Sans SC. Sizes are for 1080 lines; scale to the recording. Captions: white bold 40 px on a dark box (74 % ink) near the bottom. ## Motion Tags open; they do not fly. An edge rises, then the tag wipes open; a figure rises 24 px while it counts up on its words (0.8 to 1.2 s, easing out); a ring draws, then its leader, then its label; items and bars come one per word. Everything starts on the word that names it or on the cut that shows it. Leaving is a 0.3 s fade at the end of its line; on a cut it goes with the cut (`dur={0}`), since a fade runs into the next shot. Nothing loops or drifts: the footage moves. ## Composition and framing Place each tag in the room its shot leaves: a lower third away from the face, a figure over a wall or sky, a label beside what it rings. Keep the face and hands clear, and text 5 % in from every edge. At most two tags at once, plus a chapter tab. A tag that outlives a cut is placed again or leaves with it. Captions keep the band below 84 %. The inset is for a stretch, not a moment: once out, tags go on its ink side. ## Sound None. The recording carries the voice and its own music. ## Components - **close** — Verdict (`react/parts/verdict.tsx`): A scorecard: each point opens on its words, a teal plus or a quiet minus; the verdict is ruled off below. Fits: the pros and cons a review ends on; a recap of what was tested - **compare** — Swap (`react/parts/swap.tsx`): What one would assume, struck through, and what is true written after it in teal. Fits: a "looks the same, but" turn; a belief corrected; an old price replaced - **compare** — Versus (`react/parts/versus.tsx`): Two tags across a small "vs"; when the speaker decides, the winner takes the teal edge. Fits: two products weighed; before and after; the old way and the new - **data** — Bars (`react/parts/bars.tsx`): Two to four amounts as bars on one tag, each growing on its words; the one that matters in teal. Fits: prices or scores of a few products; a benchmark read out - **data** — Figure (`react/parts/figure.tsx`): One figure on a tag, counting up on its words, its unit in teal and what it counts above. Fits: a price, a size or a count the speaker says - **ground** — Scrim (`react/parts/scrim.tsx`): A soft shadow over one edge of the shot, so tags on a bright or busy picture stay readable. Fits: a sunny exterior under a tag; captions over a busy floor - **mark** — Callout (`react/parts/callout.tsx`): A ring drawn around something in the shot, a leader out to a short label. Fits: a sticker, a port or a button on the product in hand; a sign they mention - **open** — Chapter tab (`react/parts/chapter.tsx`): A small tab in a top corner naming the part of the video, its number in teal; it stays for the part. Fits: the sections of a review; the steps of a tutorial; an interview's questions - **quote** — Quote (`react/parts/quote.tsx`): A few words written out on a tag behind a teal quotation mark, with their source below. Fits: a line from a spec sheet, a maker or a review they cite; the sentence the video turns on - **steps** — Chips (`react/parts/chips.tsx`): Short items in a row of tags, each opening when it is named; the one being said has the teal edge. Fits: two to five things listed aloud: uses, features, reasons - **term** — Name tag (`react/parts/name-tag.tsx`): A lower-third tag naming who or what is on screen: a teal edge rises, then the tag wipes open. Fits: a person introduced; a product or model named; a place the shot shows - **title** — Headline (`react/parts/headline.tsx`): The point of a passage in a few words on a tag, the key words underlined in teal as they are said. Fits: what the speaker concludes; the question the video answers - **transition** — Inset (`react/parts/inset.tsx`): The recording draws itself into a framed window on one side, still playing; the rest becomes ink for tags. Fits: a stretch of points or a comparison beside the speaker; a recap; a shot too busy for tags ## Native moves - **Name it as they say it.** A name tag opens on the name, for the sentence about them. *Fits content like:* an expert introduced; a product model read out. - **Ring the thing in the shot.** A ring round the detail they point at, a leader to a few words. *Fits content like:* the sticker on a laptop; a port on a phone's edge. - **Count it as they say it.** A figure counts up on the words that say it. *Fits content like:* a price; a memory size; a distance run. - **Strike and correct.** What one would assume, struck, then what is true. *Fits content like:* "it looks the same, but inside it is different"; a myth corrected. - **Weigh two.** Versus or bars, the winner in teal when they say it. *Fits content like:* this laptop against last year's; three prices read out. - **Step aside for the summary.** The inset frames the recording while a verdict or list fills the ink beside it. *Fits content like:* the pros and cons at the end of a review; a recap of the three points. ## Avoid: pitfalls - A tag over the face or the hand holding the product. Fix: the other side, or the next shot. - Text the video already shows (its own price card, title, end card). Fix: nothing on a designed frame. - The caption again on a tag. Fix: a name, a figure or a few words, never the sentence. - A tag kept across a cut onto something. Fix: place it again, or let it go on the cut. - The inset in and out within a few seconds. Fix: hold it for a stretch, or use a tag. ## Freedom in each work Each video decides which moments earn a tag, which side they sit on, whether chapters are marked and how the ending is left. Three far-apart uses: a product review whose tags name, count and ring the device in hand; a travel vlog with only place names and one figure per stop; an interview where a name tag opens once and headlines mark the two answers that matter.