How to prompt Seedance

She keeps her selfie framing and speaks directly to camera with natural hand gestures: "I was skeptical, but this cleared my skin in two weeks." Slight handheld sway, hair moving softly, lighting unchanged from the source image. Ambient: quiet bedroom room tone.

In image-to-video the prompt directs only motion and the spoken line — identity, outfit and framing come from your start image. The quoted line is lip-synced natively; this exact pattern powers ManyMotions's UGC ad pipeline in production.

An amber glass skincare dropper bottle rotates slowly on a matte white turntable, centered composition, soft wraparound studio light with one specular highlight sliding across the glass, seamless light-gray background. Audio: low studio room tone, a faint mechanical whirr from the turntable.

Fixed Camera does the stabilizing, so the prompt spends every word on the product and the light. Turn the toggle on instead of writing "static camera" — the parameter is deterministic, prose is a suggestion.

The closed gift box on the table opens by itself, the lid lifting smoothly as golden light spills from inside, tissue paper unfolding, camera holding a gentle slow push-in. Audio: soft paper rustle, a warm rising chime as the light appears.

Upload the closed box as the start frame and the open, glowing box as the end frame — Seedance 1.5 interpolates a believable path between the two. Use this for reveals and transformations where the final image is non-negotiable.

A lone cyclist crosses an empty concrete overpass at dawn, tiny against the brutalist architecture, ultra-wide composition with heavy negative space, slow lateral tracking shot, cold blue-hour light with the first orange sun hitting the towers behind. Audio: distant city hum, freewheel clicks, wind.

21:9 rewards composition language — "negative space", "tiny against" — that tells the model how to use the extra width. One subject, one move, ten seconds: the 1.5 sweet spot for b-roll.

Macro shot: cold brew coffee pours over a large clear ice cube in a glass, liquid sheeting around the ice with realistic refraction, tiny bubbles rising, condensation beading on the outside, single hard backlight making the liquid glow amber. Audio: the pour, ice crackling as it splits.

Physics is a Seedance family strength, and it feeds on nouns and verbs — "sheeting", "refraction", "crackling" — not adjectives. The ice-crack foley syncs to the visible split because both are named.

A runner ties her shoes on a foggy trail at sunrise, low angle close on the hands, then she stands and runs past the camera out of frame, shallow depth of field, soft diffused light. Audio: birdsong, laces snapping, footsteps fading.

Set a fixed seed, then change exactly one phrase per run — swap "low angle close on the hands" for "overhead top-down shot" and rerun. With the seed held constant, output differences come from your wording, not randomness: prompt debugging instead of prompt gambling.

The man in the portrait blinks naturally, breathes, and slowly breaks into a warm smile, eyes catching the light, a faint breeze moving his hair, background bokeh drifting subtly.

Small verbs — "blinks", "breathes" — keep fidelity to the source photo; big verbs invite the model to reinvent the face. Pair the micro-motion prompt with Fixed Camera on so all movement belongs to the subject.

Handheld camera weaves behind a street cook flipping noodles in a flaming wok at a night market, steam and sparks lit by hanging bulbs, quick natural sway as the camera leans over the wok for a closer look. He shouts happily: "Extra chili? Extra chili!" Audio: roaring flame, clattering metal, crowd chatter.

"Handheld" plus a described camera intention ("leans over for a closer look") produces motivated movement instead of random drift. The shouted line lands with lip-sync and the wok foley times itself to the flips.

The sneaker on the pedestal rotates as the background shifts from white to deep blue, rim light tracing the silhouette, slow dolly-in from medium to close.

Rough out motion ideas on Seedance v1 Pro Fast at 480p — same prompt grammar, quicker renders, no audio. When the movement reads right, rerun the identical prompt on v1.5 Pro at 1080p with audio on.

Abstract waves of indigo and teal silk fabric ripple in slow motion, folds catching studio light, continuous fluid motion with no beginning and no end, hypnotic and calm.

Upload the same image as both start frame and end frame and the clip returns to its opening state — a seamless loop for website headers and ambient screens. Keep the motion continuous ("ripple", "flow") so the loop point disappears. Loops read cleaner silent.