The cheat sheet

Every term on one page — keep it open next to your chat. Grows as new clips ship.

Composition

Term	What it means	Say it like
`window`	Your app as captured — the input size	"capture the window at 1600 by 1000"
`output`	The file's resolution and quality	"output at 1080p, high quality"
`virtual screen`	The stage the window stands on; the camera moves across it	"give the window more room on the screen"
`background`	Solid, gradient, or image behind the window — can animate mid-shot	"animate the background to a sage gradient"
`chrome`	The window's dress: macOS, Windows, none, or a phone	"no chrome at all"

Camera

Term	What it means	Say it like
`zoom`	Point the camera — at an element, or the whole app	"zoom in on the document"
`medium`	Level 0.75 — the target plus its context	"zoom to the form, medium"
`close-up`	Level 0.95 — the target fills the frame	"close-up on the moodboard"
`super-close-up`	Level 1.20 — crops into the detail	"super close-up on the headline"
`numeric level`	The fill ratio itself: how much of the frame the target fills (1 = exactly)	"a touch tighter — 1.15"
`anchor`	The point that holds still once cropped in — nine names, or any { x, y } number	"anchor at point-seven, point-three"
`hard cut`	Instant jump, no travel	"hard cut to the whole app"
`snap`	Skips the middle of the move (half at most) — type first, snap onto the field	"start typing, then snap to the input"
`pan`	Cropped in, drift over what's in frame — can't reveal below the fold	"crop into the table, pan across it"
`follow`	The camera rides the cursor between targets — zoom unchanged	"follow my cursor to the table"
`scroll`	Travels the page to new content, smooth and on camera	"scroll down to the checklist"

Sound

Term	What it means	Say it like
`music bed`	A track you supply, laid under the whole video — starts on the first frame, loops to fit	"use dusk-loop.mp3 as the bed under the whole video"
`duck`	The bed steps down under each spoken line, then comes back	"(automatic — the bed dips under every spoken line)"
`BPM / beat grid`	Tell it the track's tempo and moments can ride the music	"the track is 94 BPM"
`change the track`	Start a second track to shift the mood — the swap waits for the bar	"crossfade to victory-lane.mp3 on the downbeat"
`stop / resume`	Drop the bed out for a quiet stretch, then bring it back	"stop the music while I type, then bring it back"
`sound effects`	Map a file you supply to clicks or typing once — every matching action carries it	"use snap.mp3 when I click the switch"
`one-off`	A sound placed by hand, not tied to an action — a chime on a reveal	"play chime.mp3 when the export finishes"
`end on beat`	The bed closes its phrase clean instead of running out mid-bar	"end on the beat"

Voiceover

Term	What it means	Say it like
`your own recording`	Lines you recorded yourself, used exactly as performed — tell your agent where the files live and when each plays	"my recordings are in ./vo — use intro.mp3 for the opening line"
`provider`	Where generated voice comes from — ElevenLabs is wired in (your API key); any other service plugs in as an adapter	"narrate it with ElevenLabs, voice Monica"
`the read`	The delivery — stage directions in brackets ([professional], [whispers]) steer a line's tone; each line is voiced on its own, so write lines that stand alone	"make this line a whisper"
`another voice`	Any line can name its own voice — two speakers in one take	"give George the second speaker's lines"
`say`	Speak a line at that moment — by default it rides over the action	"say 'the export just finished' as the toast appears"
`hold`	Keep the frame still until the line has finished	"hold on the diagram while you say it"
`subtitles`	A separate file next to the video, never burned into the pixels — on by default	"no subtitles this time"
`word timings`	The exact moment every word is spoken, available for any line — build word-by-word captions from them	"highlight each word as it's said"