Motion graphics

Sources: T13 = motion-graphics study, T22 = renderer study. Choose inside each range from this video's speech, genre and frame; say why. Captions and plain titles: bc:captions-text; transitions, punch-ins, item In/Out: bc:effects.

1. Does this sentence need a graphic?

With plan.visuals (bc:visual-plan), build its graphic rows at their at word, showing their show; question a row only when building shows it fails, and update the plan. Without one: a graphic shows what the words don't (example, number, relationship, consequence). Read transcript words: what did the speaker make the viewer imagine?

• Skip it when the card would restate the caption: captions and on-screen text take one role each, never both (openmontage). Leave some sentences bare so the next card lands (ghost-editor).
• Change the graphic when the content changes, never on a timer ("every 15 s" is an anti-pattern, digitalsamba).
• Graphics over talk keep the voice under them; never cut to a silent standalone card (vibetube).
• Vary the device; one card on every point reads as a template (T13).

2. Pick the cheapest tool that does it

1. Native text (bc:captions-text), first choice: titles, labels, keyword stickers, place and chapter cards. Editable, no render.
2. A library item: library list --panel stickers / --kind text-preset (saved card, alpha movie).
3. A graphics.render plugin (§4): kinetic type, counter, chart, clicking pointer, mock screen, full-frame card, building diagram, speaker shrinking into a card. Rendered once to a .mov with alpha; re-render when its words change. Say which and why. A graphics-led reference or genre (a dense explainer short of designed type and cards) needs §4 for most sentences.

3. Timing (anchor to words, not seconds)

Taste stated as law ("never linear", "no bounce") is not a rule: linear suits constant pans and counters; overshoot 1.05–1.12 suits playful, none calm or premium (shotcraft, saas).

Steps: payoff word's timeline frame W → entrance length E → overlay starts at W − E (the composition's entrance takes E) → hold until the narration it illustrates ends + 0.5–1 s.

4. Make it with a graphics plugin

This kit decides what, where and when; a graphics.render plugin makes it (plugins list):

• HyperFrames Graphics: read bashcut.hyperframes:hyperframes before the first graphic; its examples and catalog are raw material. Many graphics over a voice: rows in plan.cut.graphics, built in one bin/graphic plan call, never one by one, one shared render.frame for one look. Short rows: a graphic over more than ~10 s is one row per section (a growing diagram: same render.scene on each), so an edit re-renders one slice. Look before you render: bin/graphic plan --only ID --snapshot S1,S2,… shows a row as PNGs in seconds; tweak with snapshots, then one full plan (a 40 s graphic takes about 3 min to render). Brand mascot: one pose per section. A place video (real estate, a venue): all location data from one bin/osm-map call into a copy of amenity-radius; every count, minute and km on screen comes from its output.
• None ready: plugins search --capability graphics.render, install, plugins setup if its health asks. While the install waits or fails, use native text and say what is missing.
• The user's Remotion or After Effects project: they render it (ProRes 4444 or HEVC with alpha); import with library add --kind sticker --file. Never run Remotion renders for them (licence, T22).

Numbers on a card: facts table first. Before searching the web, ask open data for every figure in one call: uv run <skill_dir>/fetch_facts.py --questions facts.jsonl --json, one question per number (a Wikidata property, a World Bank indicator for a country, or a Wikipedia summary; --help for the fields). Check each matched the right place or thing. Rows go straight into widget data: value → a counter's to or a bar's value, " " + unit → unit; keep credit, date and source URL with the script line (the frame's credit only when the user asks). Answers are cached in project data facts. Search the web only for rows that came back none, never to replace an answered one; a figure the table cannot hold (a price, a fee) is a web row from the start.

Animated accents (Lottie): a check, an arrow, a loader, confetti. All of a video's in one call: uv run <skill_dir>/fetch_lottie.py --words "check,arrow,loader,confetti" --dest <project>/media/lottie draws one sheet (useAnimations MIT, Noto Animated CC BY 4.0, LottieFiles Lottie Simple License). One look, then --pick "check=2;confetti=1"; one accent = one bin/graphic place lottie (options in the plugin skill). A Noto pick's credit ("Google (Noto Emoji)") goes in the description only when the user asks. Never .lottie zips or animations with image assets.

Hand the plugin a brief, not a template: words, payoff frame, what the viewer sees, hold, look (style profile, reference frames, the video's titles).

5. Check

• Landing frame and entrance: ui frame W --phone, review window W --span E. Readable at phone size, on the word, face clear (media subjects boxes vs the card's rectangle).
• review layout for the native text around it; review shots sees the overlay as a layer.
• Re-timing the voice moves words, not the overlay: after cuts there, re-read transcript words, move or re-render.

Report

Which sentences got a graphic and why, the tool for each (native, library, rendered), its timing against the word, the render numbers (seconds, size) and what was left bare on purpose.