Playcut v2 is coming MCP · Actor v2 · & much more
Prompt library

Seedance 2.0 prompts that work around the face rule

Seedance 2.0 prompts written shot by shot, with people kept to text-to-video because image inputs with faces are rejected. Every card lists length and credits.

Updated · Playcut prompt library

ByteDance's guide for Seedance 2.0 asks for the same order in every prompt: who is doing what, then where and in what light, then how the camera moves, then style and the things to leave out. Longer clips go shot by shot, with no seconds attached.

The 30 cards below are written for Seedance 2.0, the version Playcut runs, and grouped by mode. Any card that attaches an image leaves human faces out, since Seedance refuses them, and no card comes with a sample clip.

Seedance 2.0 prompts in ByteDance's order

ByteDance's Seedance 2.0 series prompt guide on BytePlus ModelArk says to "first lock in 'who' is 'doing what'", then explain where and what atmosphere, then how to shoot, and finally tighten the result with style, image quality and constraints. Every card below keeps that order:

  1. Who is doing what, with the body part, speed and force.
  2. Where, with the light and the time of day.
  3. How to shoot: one shot size and one camera move.
  4. Sound: every Seedance render has audio, so say what it is.
  5. Style and constraints: "rich details", "keep it subtitle-free".

The guide also asks for slow, gentle, continuous movement rather than sprints and jumps, and for emotion shown through physical detail rather than words like "very sad". Keep the camera simple too: it says to "specify only 1 type of camera movement in a single shot".

Shot 1, Shot 2, Shot 3, with no seconds

For more than one beat, write Shot 1, Shot 2, Shot 3 in the order events happen, and leave durations out. ByteDance's tip: "The model's support for precise timing (such as 0–3 seconds) is unstable, and forcibly limiting duration may lead to abnormal generation results."

Constraints go inside the prompt

Playcut does not pass a separate negative prompt to Seedance, so exclusions belong in the prompt. Use the guide's own phrases, such as "keep it subtitle-free", "do not generate a logo" and "do not generate a watermark".

The face rule, and what to attach instead

Seedance 2.0 rejects any input image that shows a human face: the Animate first frame, every reference image and both Transition frames. AI-generated faces count too, including portraits made on Playcut. The job fails with InputImageSensitiveContentDetected.PrivacyInformation and the held credits are refunded.

The limit is ByteDance's: its Seedance 2.0 series tutorial says the models do not support reference images or videos containing real human faces. Text to video is unaffected, so write people in words and attach products, rooms, streets, landscapes or illustrations without human faces. For a person from a photo, switch to HappyHorse 1.1 or Wan 2.7, which accept photos of people.

Seedance 2.0 limits in Playcut

Setting Seedance 2.0 on Playcut
Modes Describe (text to video), Animate (first frame), Reference, Transition (first and last frame). No Extend and no Edit.
Aspect ratios 21:9, 16:9, 4:3, 1:1, 3:4, 9:16. The default is 16:9.
Resolution 480p, 720p, 1080p, 4K.
Length 4 to 15 seconds, whole seconds only. The default is 5.
References Up to 9 images, plus one audio reference of 15 seconds or less, in Reference mode only.
Mixing inputs A first frame and reference images cannot go in the same request.
Sound Always on. Effects and ambience are strongest; dialogue lip-sync is best-effort.
Credits per second 10 at 480p, 21 at 720p, 60 at 1080p, 108 at 4K.

A 5-second clip costs 50 credits at 480p, 105 at 720p, 300 at 1080p and 540 at 4K; the Seedance 2.0 limits and credit rates page has the full table. Use the aspect ratio calculator to check a 21:9 or 3:4 frame before you pick one.

People scenes, written in text only

Run on Seedance 2.0

Text to video is the one Seedance 2.0 mode that takes people, because no image goes in. Lines go in double quotes after "says:", in one language per clip, and lip-sync is best-effort. Cards are 16:9 unless the title names a ratio.

  1. Street dancer under a railway bridge (9:16)

    8 s · 720p · 168 cr
    A street dancer in a yellow windbreaker dances alone under a railway bridge at night, rolling her shoulders slowly, then sliding one foot out in a smooth arc. Wet cobbles reflect orange lamps. Fixed medium shot. Sneakers squeaking, a distant train. Keep it subtitle-free.
  2. Chef calling the pass

    6 s · 720p · 126 cr
    A chef in a white jacket slides a plate onto the pass, taps the bell twice and says: "Table six, walking." Copper pans glow under the heat lamps behind her. Slow push-in to a close-up. Pans sizzling. Keep it subtitle-free.
  3. Violinist rehearsing in an empty hall

    10 s · 720p · 210 cr
    A violinist in a grey sweater rehearses alone on a bare stage, bow moving slowly, shoulders loosening as the phrase resolves. Empty red seats fade into darkness. Smooth lateral tracking shot from the wings. One sustained violin melody, long hall reverb. No text on screen.
  4. Two friends under festival lanterns

    8 s · 720p · 168 cr
    Two friends walk slowly down a street hung with paper lanterns on a festival night. One stops, points at a fish-shaped lantern and says: "That one is ours." Smooth tracking shot from behind. Distant drums, crowd murmur. Keep it subtitle-free.
  5. Greenhouse gardener, with voiceover

    10 s · 1080p · 600 cr
    An older gardener in a canvas apron mists seedling trays in a glass greenhouse at dawn, moving slowly along the bench. Soft light through fogged glass. Fixed wide shot. Voiceover: A calm female voice says: "Everything here started smaller than a grain of rice." Avoid generating any text or subtitles.
  6. Grandfather and boy at a koi pond

    6 s · 720p · 126 cr
    A grandfather crouches by a koi pond and shows a small boy how to scatter food, opening his palm slowly. The boy leans in and whispers: "They are coming." Orange koi rise to the surface. Low close-up from the water's edge. Water lapping. No subtitles.

Multi-shot storyboards: Shot 1, Shot 2, Shot 3

Run on Seedance 2.0

Shots go in the order events happen, with no seconds, because ByteDance's guide calls precise timing unstable on Seedance 2.0. Give each shot one camera move, and allow 10 to 15 seconds for three shots.

Plan the shots before you write them with the shot list generator

  1. Desert highway, noon to dusk

    15 s · 720p · 315 cr
    Shot 1: Fixed wide shot of an empty desert highway at noon, heat haze over the asphalt. Shot 2: A dusty green camper van enters from the left; slow pan to follow it. Shot 3: Close-up on its tail lights shrinking toward red hills at dusk. Wind, a fading engine. No subtitles.
  2. Cargo bike morning delivery

    12 s · 720p · 252 cr
    Shot 1: A grocer loads crates of oranges into an electric cargo bike outside a shuttered market; fixed shot. Shot 2: Tracking shot beside the bike along a canal at sunrise. Shot 3: Fixed medium shot as she sets a crate down at a café door and says: "Fresh this morning." No subtitles.
  3. Glassblower's first bowl

    12 s · 720p · 252 cr
    Shot 1: Close-up of a glowing orange gather turning on the end of a blowpipe; fixed shot. Shot 2: The glassblower blows gently as the bubble swells; slow push-in. Shot 3: Fixed wide shot of the finished amber bowl cooling on a steel bench. Furnace roar, then the tick of cooling glass.
  4. Morning surf check

    10 s · 720p · 210 cr
    Shot 1: Aerial view of grey dawn waves folding onto an empty beach; slow forward drift. Shot 2: A surfer waxes a board on the sand, breath fogging; fixed medium shot. Shot 3: The surfer paddles out through white water; tracking shot from behind. Crashing surf, gulls. No logo.
  5. Mountain railway in snow

    15 s · 720p · 315 cr
    Shot 1: Drone shot following a red train as it winds along a snowy mountainside. Shot 2: A gloved hand wipes a fogged carriage window to reveal white peaks; fixed close-up. Shot 3: Fixed wide shot as the train slides into a tunnel. Wheels clacking, a long horn. Keep it subtitle-free.

Product and object references, no faces

Run on Seedance 2.0

Reference mode takes up to nine images, but ByteDance recommends four or five assets, so none of these uses more than four. Name each image beside its subject, as in "the white canvas sneaker from Image 1". An unnamed image is effectively ignored, and faces are rejected.

Make a faceless product still to use as Image 1

  1. Sneaker on a skate ledge (3:4)

    8 s · 720p · 168 cr
    The white canvas sneaker from Image 1 rests on the concrete skate ledge from Image 2 in late afternoon sun. Slow orbit around the sneaker, stitching and toe cap sharp. An empty skateboard rolls past behind it, wheels clacking. Grainy film texture.
  2. Pour-over kettle on a slate counter

    6 s · 720p · 126 cr
    The ceramic pour-over kettle from Image 1 pours a thin stream into the glass dripper from Image 2 on the slate counter from Image 3. Fixed close-up at counter height. Steam curls as the coffee drips. Water trickling. Morning window light, no subtitles.
  3. Cargo bike in your brand's colours

    5 s · 720p · 105 cr
    The electric cargo bike from Image 1 stands on the cobbled street from Image 2, its front box full of potted herbs, as morning light slides across it. Grade the colours to the palette card in Image 3. Slow push-in at wheel height. Birdsong. No text, no watermark.
  4. Fox mascot in a storybook forest (1:1)

    8 s · 720p · 168 cr
    The illustrated fox mascot from Image 1 trots into the storybook forest from Image 2 and stops to sniff a glowing mushroom. Keep the fox's flat colours and thick outlines exactly as drawn. Fixed wide shot. Wind chimes, rustling leaves. 2D animation, no text.
  5. Greenhouse rain, with an audio reference

    6 s · 720p · 126 cr
    Rain streams down the glass roof of the greenhouse from Image 1 while the ferns inside sway slightly. Slow push-in toward a dripping leaf. Use the rain and thunder in Audio 1 as the soundtrack. Grey afternoon light, keep it subtitle-free.

    Attach one rain recording of 15 seconds or less as the audio reference. Audio references work in Reference mode only.

  6. Picnic basket, lit like a photo

    5 s · 1080p · 300 cr
    The wicker picnic basket from Image 1 sits on the checked blanket in the meadow from Image 2 as long grass sways around it. Light it like the golden-hour photo in Image 3. Slow push-in. Bees, a skylark overhead. No added text.

Start from a first frame (image to video)

Run on Seedance 2.0

Animate uses one faceless still as the opening frame, so the prompt needs only motion, one camera move and sound. A first frame cannot be combined with reference images in the same request.

  1. Koi pond still

    6 s · 720p · 126 cr
    The koi glide slowly under the surface and one breaks it with a small ripple. Lily pads rock gently. Slow push-in toward the ripple. Trickling water, a distant bird. No subtitles.
  2. Camping lantern on a rock

    5 s · 720p · 105 cr
    The lantern's flame flickers on and steadies, throwing warm light over the rock and the pine needles as dusk deepens. Fixed shot. Crickets start up, wind in the trees. No logo, no watermark.
  3. Illustrated skyline (21:9)

    8 s · 720p · 168 cr
    Lights switch on window by window across the illustrated skyline as the sky turns from orange to deep blue. A small plane blinks across the top. Slow pan from left to right. A soft synth pad. Keep the flat illustration style.
  4. Empty lantern street at dusk

    8 s · 720p · 168 cr
    The paper lanterns above the empty street sway in a light breeze and brighten one by one as dusk falls. Petals drift over the stones. Slow dolly forward. Distant festival drums, wind chimes. Keep it subtitle-free.
  5. Empty lido at sunrise

    6 s · 720p · 126 cr
    Sunrise spreads across the empty outdoor pool as a light wind ripples the water and the painted lane lines waver beneath it. Steam lifts off the surface. Fixed wide shot. Birds, a flag rope tapping its pole. No subtitles.

First and last frame transitions

Run on Seedance 2.0

Transition takes exactly two faceless frames, a first and a last, and no references; the prompt describes what happens between them. Any whole-second length from 4 to 15 works.

  1. Shoebox to sneaker reveal

    5 s · 720p · 105 cr
    The lid of the kraft shoebox lifts off and tissue paper unfolds as the white canvas sneaker rises and settles on the lid, ending on the last frame. Fixed shot. Paper rustling, a soft thump. No added text.
  2. Greenhouse from day to night

    8 s · 720p · 168 cr
    Daylight fades through the glass greenhouse and string lights flicker on along the roof beams, moving from the bright first frame to the glowing last. Fixed wide shot. Birdsong gives way to crickets. No subtitles.
  3. Pencil sketch to finished frame

    6 s · 720p · 126 cr
    The pencil sketch of the desert highway fills with colour from the horizon outward, lines becoming asphalt, rock and sky until it matches the photo in the last frame. Slow push-in. Pencil scratching, then desert wind. No subtitles.
  4. Empty dripper to full carafe

    6 s · 720p · 126 cr
    Coffee drips into the glass carafe, rising from empty in the first frame to full in the last as the steam thickens. Fixed close-up. Steady dripping, a soft gurgle at the end. Warm morning light.

Draft at 480p, finish at 4K

Run on Seedance 2.0

Seedance 2.0 is the cheapest video model we run per second at 480p, at 10 credits. Draft at 480p until the prompt behaves, then run it again at 4K: that is a new render, not an upscale of the draft. Each tag shows both 5-second costs, 50 and 540 credits.

  1. Redwood fog (21:9)

    5 s · 50 → 540 cr
    Morning fog rolls between towering redwood trunks as sunbeams cut through in long diagonal shafts. Slow lateral tracking shot at ground level, ferns in the foreground. Dripping water, a distant woodpecker. Rich details, no subtitles.
  2. Soap bubble drifting, macro

    5 s · 50 → 540 cr
    Macro shot of a soap bubble drifting through still air, rainbow swirls sliding across its surface against a dark background. Fixed shot. A faint breath, then a soft pop. Rich details, no text, no watermark.
  3. Autumn leaves on a mountain stream

    5 s · 50 → 540 cr
    Red and gold maple leaves drift down a clear mountain stream, spinning slowly over smooth stones. Low tracking shot just above the water. Running water, wind in the trees. Rich details, keep it subtitle-free.
  4. Terraced rice fields from the air

    5 s · 50 → 540 cr
    Aerial shot drifting over terraced rice fields full of water that mirrors the sunrise, mist in the lower terraces. Wind, distant frogs. Rich details, no text.

Setting up a Seedance 2.0 run in Playcut

After you paste a card into the video studio:

  1. Pick Seedance 2.0 in the video model picker. Veo 3.1 is the default.
  2. Choose the mode the group names: Describe, Animate, Reference or Transition.
  3. Set the ratio, resolution and length shown on the card's tag.
  4. Attach images in the order the prompt numbers them, so Image 1 goes in first.

Long renders can hit Playcut's 9-minute limit, most often at 15 seconds and 1080p, so draft short and low first. If a job fails, the credits held for it come back.

The camera movement prompt generator helps you pick one camera move per shot.

FAQ

How do I reference an image in a Seedance 2.0 prompt?

Name each attached image beside the subject it shows, in upload order: "the white canvas sneaker from Image 1". ByteDance's guide writes the label as Image 1 and Playcut's API docs write @Image1; either way, an image the prompt never names is effectively ignored.

Why did Seedance 2.0 reject my image?

The error InputImageSensitiveContentDetected.PrivacyInformation means the image shows, or may show, a person. Seedance 2.0 rejects faces in first frames, reference images and transition frames, AI-generated ones included, and the held credits are refunded. Use a faceless image, or write the person into a text-to-video prompt.

How do I write a multi-shot Seedance prompt?

Write Shot 1, Shot 2 and Shot 3 in the order events happen, each with one camera move, the action and the sound, and leave the seconds out. ByteDance's guide warns that precise timing is unstable on Seedance 2.0. Allow 10 to 15 seconds for three shots.

Can Seedance 2.0 speak dialogue?

Yes, as a best effort. Every render has sound, and a line in double quotes after "says:" is voiced, but lip-sync varies; effects and ambience are its strongest audio. Keep lines short and in one language per clip, as ByteDance's guide advises.

How long can a Seedance 2.0 clip be?

From 4 to 15 whole seconds at every resolution from 480p to 4K; Playcut's default is 5. Credits are per second, so 15 seconds costs 315 credits at 720p and 1,620 at 4K.