Brief the director, not the render.

"A silver-haired watchmaker in a navy apron and round glasses assembles a tiny gear under a desk lamp"

Three stable features (hair, apron, glasses) become the lock. Vague subjects (“a man works”) force the model to invent, and re-invent, a face every shot.

"A matte-black ceramic pour-over coffee set with a walnut collar, steam rising, on a rain-lit windowsill"

Material + color + one distinguishing detail (walnut collar) give the renderer something to keep constant while the camera moves.

"A tall courier in a yellow rain jacket hands a package to a short baker in a flour-dusted apron, at the bakery door at dawn"

Each subject gets its own visual tag. Untagged pairs merge into each other mid-video, the classic two-person artifact.

"A climber chalks her hands slowly, exhales, then commits to the first hold of an overhanging gym wall"

Three sequential micro-actions with natural inertia. The engine paces them across beats; “an epic climbing video” gives it nothing to choreograph.

"A student opens an exam result on her phone; her shoulders drop, then she laughs and spins her chair once"

Relief written as behavior. Abstract words (“she is overjoyed”) under-perform, models act what they can see.

"A detective in a long coat stalks down a neon-lit arcade alley in the rain"

“Stalks” vs “walks” vs “trudges” changes the whole performance, one verb swap, no other change needed.

"Our new running shoe launches next month, make people feel it's inevitable"

Emotion words (“inevitable”) steer Auto to Tentpole with a Mythic tint: low hero angles, slow push-ins, an 85mm payoff. You describe the feeling; the engine builds the grammar.

"An empty lighthouse kitchen at night; a kettle begins to whistle and no one comes"

Atmospheric Tension holds wides longer, drops the key light, and lets the whistle escalate, slow is powerful when tension rises, so don't ask for “fast cuts”.

"A cat sits on a robot vacuum, dignified; the vacuum starts and slowly rotates the cat out of the room"

Comedy Timing shoots this flat, wide and locked, holding the frame one beat too long. The mode carries the joke's rhythm.

"A sneaker designer sketches, stitches and laces one shoe, every gesture landing on the beat"

Musical Rhythm cuts on accents and freezes on the final beat, say what should land on the beat, not where to cut.

"An empty concrete rooftop becomes a lantern-lit rooftop cinema as the sun sets"

Before (empty, daylight) and after (glowing, night) give every development beat a direction to travel.

"A dusty garage door rolls up to reveal a restored 1970s motorcycle, chrome catching the morning light"

One concealment + one reveal = a natural hook and payoff. The engine holds the reveal for the final beat instead of spending it on shot one.

"A potter shapes a bowl through one long afternoon; the light moves across the studio as the bowl finds its form"

Minimalist Art House reads “time passing” as the event itself, small change, long takes, no forced drama.

"For TikTok, 9 seconds: a barista snaps a lid on a cup, slides it down the counter, and it stops exactly on the logo coaster"

Short duration + platform = the engine compresses to a 3-beat arc and front-loads the strongest image.

"For YouTube, 15 seconds: a family tailor shop through three generations, same doorway, same bell, different hands"

Longer canvas → the engine gives development beats room and lands an emotional payoff instead of rushing the cut.

"For Instagram, 8 seconds: raking light crosses a hand-stitched leather wallet, thread by thread, ending on the maker's stamp"

Macro detail + one light move + a stamp payoff, Architectural Precision keeps every frame composed before motion.

"Using my saved brand: a cyclist crests a hill at dawn and coasts into the city as the streetlights blink off"

No need to write hex codes or logo instructions, the BRAND_BLOCK is locked and appended separately, never mangled into the scene text.

"Make a 12-second launch ad from https://yourstore.com/products/aurora-lamp"

The engine scrapes an editable draft, locks the product hero image as the continuity reference, and directs around the real product.

"Two friends on a night bus; one looks up from her phone and says “we could just go now”, and the other smiles"

Quoted speech routes to a dialogue-strong model; one short line lands, monologues drift.

"A red paper boat sails through a city where the streets are slow rivers of fog, always the same red boat"

Surreal worlds can morph, the repeated anchor (“always the same red boat”) is what keeps the dream coherent.

"A rice farmer in Chitwan checks the sky, then her phone's weather app, and smiles at the forecast"

Documentary Truth uses available light and observational handheld, asking for “cinematic drama” here would fight the mode. Real beats staged.

"Remix the “rooftop cinema” direction from the gallery, but for our bookshop's late-night reading club"

Remix carries the mode + shot skeleton; your subject replaces theirs. Craft is shared, brands never leak.

"For TikTok, 12 seconds, using my saved brand: a courier in a yellow jacket races dawn traffic to deliver a birthday cake intact, handheld energy, and the payoff is the kid's face, not the cake"

Subject features + stakes + platform + duration + brand + where the emotion lands. One sentence, everything the director needs, this is the ceiling of the craft.