Creative fundamentals

The decisions AI does not make for you: what the viewer sees first, the story, the pace, colour, type and brand consistency.

Also called hooks, scroll stopper, first second, thumb-stop

What: Whatever stops someone scrolling: the first one to three seconds of a video, or the single element that reads instantly in an image.
Why: Nothing else in the piece matters if people never see it. A better hook often beats a better middle.
Example: Open on the cracked phone screen and the words "Dropped it again?" instead of a slow logo reveal.

Watch out

Opening with a logo or a slow establishing shot.

Also called problem, product, proof, CTA, call to action, CTA, script structure

What: A simple shape for the message: problem, product, proof, call to action for ads; what it is, how it works, why it matters for explainers.
Why: Writing the story in plain sentences first makes every generation purposeful and the edit straightforward.
Example: Problem: oily skin by noon. Product: the gel. Proof: before and after at 4 pm. CTA: shop now.

Watch out

Generating clips first and hunting for a story in them afterwards.

Also called wide shot, medium shot, close-up, camera movement

What: Wide shots show where we are, medium shots show a person and action, close-ups show detail or emotion. Camera moves include pan, tilt, dolly and handheld.
Why: Naming shots precisely gives better prompts and a shot list the client can approve.

Watch out

Every clip a wide, slow drone-style shot, so nothing feels close or urgent.

Also called pacing, cuts, editing rhythm, cut

What: How long each shot stays on screen and why you cut. Short ads often cut every one to two seconds; explainers breathe longer. Cut when a thought is complete.
Why: AI clips are short and imperfect: choosing the best seconds and cutting to a rhythm is most of the craft.

Watch out

Letting every generated clip play to the end because it cost credits.

Also called voiceover, music, sound effects, audio mix

What: Voiceover, music and small sound effects, mixed so the voice sits clearly above the music.
Why: Sound makes cheap visuals feel finished; muddy audio makes good visuals feel amateur.

Watch out

Music so loud the voiceover is hard to follow on phone speakers.

Also called colour harmony, HEX code, contrast ratio, colour grading, palette

What: Palettes built from harmonies (complementary, analogous, triadic), brand colours given as HEX codes, contrast that keeps text readable, and grading that makes clips match.
Why: Colour sets mood, guides the eye and is the fastest way a brand is recognised.

Watch out

Checking colours only on a good laptop; test on a cheap phone in daylight.

Also called typeface, font pairing, leading, kerning, text on images

What: How text looks: typeface, size, weight, leading (line spacing) and kerning (letter spacing). Pair at most two typefaces and keep text short and high contrast.
Why: Text is where clients look first and where AI errors hide, so most designers add final text by hand as editable layers.

Watch out

A misspelled brand name baked into a generated image.

Also called on-brand, consistency across assets, brand rules

What: Using the same colours, fonts, logo placement, tone and image style in every asset, checked against the brand kit.
Why: Clients value a set that looks like one company over any single lovely asset.

Watch out

Ten great assets that look like they came from ten different brands.

Go deeper