How to prompt Seedance 2.0

A woman in her 30s in a bright bathroom holds a green glass serum bottle up to the camera, selfie framing, slight handheld sway, morning window light. She says with a casual smile: "Okay, one week with this and my skin is actually glowing — look at this." She tilts the bottle so the label catches the light. Ambient: quiet room tone, faint birdsong outside.

The quoted line is lip-synced natively — no separate TTS pass. Selfie framing + handheld sway reads as authentic UGC; naming the label moment forces a product beat instead of generic waving.

A matte black wireless earbud case rotates slowly on a marble pedestal, studio product photography, locked-off static camera, single softbox from the left with a rim light from behind, dark seamless background, macro detail on the hinge as it opens. Audio: soft mechanical click when the lid opens, low ambient hum.

"Locked-off static camera" replaces the missing camera-lock switch — say it explicitly or the model will drift. One named micro-event (the lid opening) gives the clip a beat and a foley cue to sync.

Neon-lit pizzeria on a rainy street corner at night, slow lateral tracking shot past the fogged window, a cook tosses dough inside, reflections smearing across wet asphalt, shot on 35mm, anamorphic lens flares, moody teal-and-red palette. Audio: rain, muffled Italian radio from inside, a passing car.

Tracking + a single interior action gives layered parallax. The palette and lens words are followed closely — this model rewards specific color direction over adjectives like "beautiful".

Shot 1 (0-4s): wide establishing, a hiker reaches a foggy ridge at dawn, drone orbit. Cut to Shot 2 (4-8s): close-up of her boots crunching frost, low tracking shot. Cut to Shot 3 (8-12s): she turns to camera and says "worth every step", handheld, golden light breaking through. Continuous ambient: wind, distant birds; boots foley in shot 2.

Numbered shots with timings produce real cuts inside a single clip — no external editing. Keep one subject across shots and restate her ("the hiker", "she") so identity holds.

A ceramic mug slides off a wooden desk in slow motion and shatters on a concrete floor, fragments scattering with realistic weight, locked-off low-angle camera at floor level, hard window light casting long shadows. Audio: the slide, a beat of silence, then the sharp shatter with room echo.

Seedance 2.0 is strong on collisions and material behavior — name the materials (ceramic, concrete) and the physics you expect. The "beat of silence" line times the audio to the fall.

First-person POV: gloved hands push open a heavy vault door, flashlight beam sweeping over dusty shelves. Mid-clip, switch to a reverse shot: the explorer framed in the doorway, backlit by the corridor. Audio: metal groan of the door, footsteps with stone echo, tense low drone.

"Switch to a reverse shot" is understood as an in-clip cut — one of the camera moves this model handles that most others refuse. Keep the location constant so the reverse shot matches.

Two friends at a diner booth, over-the-shoulder framing that alternates with each line. The first, a man in a denim jacket, says: "You actually quit?" The second, a woman stirring coffee, smiles and answers: "Signed the papers this morning." Warm practical lamps, soft film grain. Audio: cutlery, low diner chatter under the dialogue.

Attribute each quoted line to a described speaker and the lip-sync lands on the right face. Alternating over-the-shoulder framing is a camera instruction, not an edit note — the model stages it.

Extreme close-up of hands cracking an egg one-handed over a sizzling pan, whip-pan up to a chef grinning at camera, kitchen towel over shoulder, punchy overhead light. He says: "Stop cracking eggs like a tourist." Audio: loud sizzle, the crack, a short record-scratch on the whip-pan.

Front-load the visual hook in the first second — the close-up action — then the whip-pan delivers the face and the line. Sound effects described at exact moments become the edit rhythm.

Bring this photo to life: the model keeps her pose but turns her head slowly toward camera, hair moving in a light breeze, background bokeh shimmering, subtle parallax as the camera pushes in a few centimeters. Audio: soft wind, distant city ambience.

In image-to-video the prompt describes only the MOTION — composition, wardrobe and light come from your uploaded frame. Small camera moves ("a few centimeters") keep the source photo's fidelity.

Using the reference images of the astronaut character: Shot 1 — she runs across a red desert plain, wide tracking shot, dust kicking up. Shot 2 — close-up inside her helmet, eyes scanning, HUD reflections. Shot 3 — she plants a flag, low heroic angle, sun flare. Same suit, same face throughout. Audio: breathing inside helmet, wind, a swelling synth note on the flag plant.

Reference-to-video accepts up to 9 reference images and holds the character across every shot — the fix for identity drift in multi-scene stories. Restate "same suit, same face" as cheap insurance.