Everything is video

And every layer of it is yours to specify.

There is almost no ceiling on what can be produced. Motion graphics, voice, music, captions, thumbnails, both aspect ratios, any number of scenes — the pipeline will build it. The ceiling is how precisely you can describe what you want. That is the whole constraint, and it is the thing this seat is actually judged on.

A social post is not a caption with a picture. It is a rendered video with a thumbnail, a voiceover, burned-in captions, motion, a music bed and two aspect ratios — and every one of those is a decision you make, or a decision someone else makes for you.

The layers, and how each is got wrong

Layer	What you specify	Specific enough	The usual failure
The composition	Which Remotion composition, how many scenes, how long each holds.	“SceneReel, 6 scenes, 900 frames at 30fps — scene 3 holds 5s, the rest 4s.”	“About 30 seconds.” The renderer cannot act on that.
The focal element, per scene	One thing the eye goes to in each frame, and what it does.	“Scene 2: the temperature reading, centred, counting up to 18°F then holding.”	Two focal points in one frame. Reads as neither.
Motion	Which sanctioned move, and where it starts and stops.	“Convergence on scene 2 only. Nothing moves during the price card.”	“Add some nice animation.” Unbriefable, unreviewable.
On-screen text	Every word that appears, exactly as it appears.	“Seven checks. About 70 minutes.” — scene 1, lower third, 2s in.	Leaving captions to whoever renders it. On-screen text is copy, and the voice gate checks it.
Voiceover	The script verbatim, plus pace, register and where it breathes.	“Calm, conversational, unhurried — a neighbour on a driveway. Pause before the price. No lift at the CTA.”	Writing the words and leaving the read to chance. The same sentence sells or reassures depending on delivery.
Captions	Burned-in styling, position, and whether they mirror the VO exactly.	“Burned in, bottom third, safe area above the platform UI. Verbatim.”	Forgetting that most viewers watch muted. Captions are the primary channel, not an accessibility afterthought.
Background music	The bed, its level under the voice, and what it must not do.	“Light acoustic, well under the read, no build and no swell at the CTA.”	A track that crescendos into the call to action — the single fastest way to make honest copy sound like an advert.
The thumbnail	The still that decides whether any of the above is ever seen.	“Technician mid-check, report visible, no text overlay.”	Treating it as an export setting. It is the most-viewed frame you will ever make.
Both aspect ratios	16:9 and 9:16 from one source, and what survives the crop.	“Scene 2 is the widest — check the reading is still legible vertical.”	Briefing only the horizontal. The vertical cut is where the point gets cropped out.
Audio master	Levels, so it does not arrive quieter than everything around it.	—16 LUFS, consistent across the week.	Not yours to execute, but yours to notice when it is wrong.

Exhaust the capability

Most people brief at about thirty per cent of what the system can do, because they are used to tooling that punishes ambition. Here the render cost of an elaborate scene and a plain one is roughly the same. Under-briefing is not caution — it is leaving the work unfinished and calling it a style.

Specify every layer, every time. A brief silent on music gets whatever was lying around.
Write the words that appear on screen. All of them, exactly.
Describe motion as a start and a stop, not as a feeling.
Say what must NOT happen. Half of a good brief is the forbidden list — it is the half that survives contact with someone else’s taste.
Then look at what came back and refuse it if it is wrong. You have sign-off. A render that is technically perfect and works against the argument does not ship.

The one line worth keeping: the machine will build anything you can describe and nothing you cannot. Your job is to leave it nothing to guess at. See how we storyboard for the format that carries all of this.