Video Essayist
Cuts a script-led argument where narration drives the timeline and every visual proves, illustrates or contrasts the line being spoken.
This is the brief the Video Essayist gives when the user tags @video-essayist or the planner activates this skill. It is professional advice inside the user's brief: the user's goal, explicit exclusions, brand rules, access and spending limits win. Its numbers are proposed targets; check that each applies and can be measured before using it. For an evidence-grounded shot plan from this specialist, use the specialist-consultation workflow (agentWorkspace action specialist_consult) instead of reading the brief alone.
Work the DERIVE list out first, from the clips actually on the timeline, and say the figures you derived. Before calling a step done, check it against the ACCEPT line. Sources, and the fair-use notes (not legal advice): references/sources.md. The full long-form brief: references/craft.md (load it when planning a whole edit in this style).
Brief
INTENT: the argument leads; every image earns its place by proving a line. SPEC: narration 140 to 160 wpm (110 to 130 for dense passages, about 170 at most). Runtime is words over pace plus 10 to 20% (1,500 words is about 10 min at 150 wpm). Cold open 10 to 30s; thesis by 60s. A new visual every 3 to 8s with breathing gaps. Borrowed clips usually 2 to 10s, longer only when the clip is what is being critiqued. Quotes held for words divided by 3 plus 1 second, with author, work and year. Chapters 00:00, at least 3, each at least 10s. Music at least 8 dB under the voice; about -14 LUFS. DERIVE FIRST: · Script word count and the measured narration pace. · Argument beats (chapters) and the visual each claim line needs. · Each borrowed clip's source, length and whether it is itself the subject. · Tone shifts, and claims that need an on-screen citation. CRAFT · Lock the narration first; then pick visuals to match. · Show, then explain: play, freeze or slow, narrate over the still. · Make critique visible: annotations, side-by-sides, freeze frames. · Hold faces before and after the key line. · Photos slide in with easing, not a uniform zoom; uncluttered maps. · Quotes and citations with addTitle; music ducked (autoDuck) and changed on tone shifts; setChapters. DO NOT: wall-to-wall narration; b-roll that matches words but not the claim; long unexamined clips; "no copyright intended" as a defence; one cue through every tone; a thesis buried past the first minutes. ACCEPT: each sentence has a visual within half a second of its key word, pace 120 to 170 wpm, thesis by 60s, valid chapters, every borrowed clip annotated, labelled and no longer than its point, music at least 8 dB under the voice.
Gotchas
- Fair use is decided case by case (US law); short duration alone does not make a clip safe, and other countries use different exceptions. Say so rather than promising a clip is "safe".
- A voice line cut through a word blocks finishing; rewrite the narration shorter instead of trimming it.
- Archival photos look dead with no motion and fake with a uniform zoom; vary direction and speed per photo.