Pricing
FLAGSHIPByteDance · Seedance 2.5

Shoot a 30-second take with Seedance 2.5

You stop stitching clips into a sequence. Seedance 2.5 holds one take up to 30 seconds. Sound lands in the same pass.

Free account. ⚡50 on signup. No card.

Official showcase

The Seedance 2.5 launch films

Every clip in this section is an official ByteDance Seedance 2.5 showcase video, mirrored as published. Prompts are reproduced verbatim where ByteDance published them.

The Morning Bakery
Text to Video

The morning bakery

Photoreal hands, paper and pastry. Texture the camera can sell.

Panda Cub in Morning Light
Text to Video

Panda cub, morning light

Fur, grass and backlight hold at 720p. No uncanny drift.

Brand Demo: Airport Scanner
Text to Video

The scanner brand film

A branded case rides the belt. On-frame text stays legible.

White Model to Underwater Fable
White-model control

Blocking in, film out

Cones and cubes set the camera. The render keeps it.

What's new in Seedance 2.5

01

The take got twice as long

ByteDance extends single-pass generation from 15 to 30 seconds. Seedance 2.0 stopped at 4–15s.

02

References in one pass

Feed up to 30 images in one pass. Plus 10 video clips and 10 audio clips. The model tracks each one across the take.

03

Edit by timestamp

ByteDance added precise editing via timestamps, per the release note.

04

Keep going past one take

Multi-round extension appends shots to existing outputs. Characters, environments and pacing stay consistent.

30 seconds. One take.

Whole scenes, not fragments

Official ByteDance Seedance 2.5 narrative showcases. Prompts are omitted in this section. ByteDance publishes only a truncated preview.

The Barbershop's Last Night
Text to Video · 28s · dialogue

The barbershop's last night

Two characters, five beats, one goodbye. Dialogue lands on frame.

Peking Opera, One Take
Text to Video · 20s · one take

Peking opera, one take

The camera circles three performers. Nothing cuts, nothing slips.

Lantern Festival in Lin'an
Text to Video · 30s

A festival with a cast of dozens

Crowds, lanterns and a verse. Thirty seconds, zero stitches.

Late-Night Pharmacy, Paris
Text to Video · 30s · French dialogue

The pharmacy at midnight

Three lines of French. Rain, neon and restraint.

New Wave at the Seaside
Text to Video · 25s · dialogue

A conversation in six beats

Two people, one bench, real timing. The pauses survive.

Seedance 2.5 specs

StatusLive in the composer
Resolutions480p, 720p
Aspect ratios16:9, 9:16, 1:1
Clip length4–30s
AudioNative audio
Reference imagesUp to 9
Cost on OpenClipsFrom ⚡85
ByteDance model limitUp to 30 images, 10 video clips and 10 audio clips per pass
ExtensionMulti-round extension appends shots to existing outputs.
Timed editingTimestamps steer camera, movement and rhythm inside a chosen time frame
Known limitsByteDance notes complex motion physics still has room to improve

Steer it with references

Show it the shot. It matches the shot.

Official ByteDance demos of video edit, sound edit, green screen, white-model control and extension. The input sits in the frame above its output.

Video Edit: Coffee to Vineyard
Video edit

One prompt re-shoots the scene

Verbatim prompt
[Video Edit]Change the background to a vineyard in France, replace the coffee in the two people’s hands with red wine glasses, change the girl’s outfit into a French floral wrap dress, and add a thin-strap small bag on her body. Keep everything else unchanged.
Sound Edit: New Narration
Sound edit

Edit the audio, keep the picture

Verbatim prompt
[Sound Edit] Remove the background music from this video and add narration that fits the visuals.
Green Screen to Fairy Tale
Green screen

The plate becomes a forest

Previz to Chase Scene
White-model control

Previz to blockbuster

Storyboard to Disaster Film
Storyboard control

Nine panels, one film

Car Render: White Model to Paint
Render pass

Paint the white model

Paper Puppy Meets Real Puppy
Reference to Video

A 2D character in a 3D world

Beat-Synced Four-Idol Film
Reference to Video · audio ref

Four faces, one track, on the beat

Video Extension: The Kitchen Scene
Video extension

The take keeps going

Seedance 2.5 vs Seedance 2.0

One long take or 4K.

Seedance 2.5Seedance 2.0
Resolutions480p · 720p480p · 720p · 1080p · 4K
Clip length4–30s4–15s
AudioNative audioNative audio
Reference imagesUp to 9Up to 9
Aspect ratios16:9 9:16 1:116:9 9:16 1:1
Cost on OpenClipsfrom ⚡85from ⚡56

Need 4K on delivery? Render on Seedance 2.0. Need 1080p in ~2–3 min with tight camera control? That's Seedance 1.5 Pro. Every version sits on one wallet, so you can draft on one and finish on another.

Credits

What a Seedance 2.5 render costs

Seedance 2.5

From ⚡85. 480p ⚡85 for 4s. 720p ⚡189 for 4s.

Video is paid per render, always. A 4-second 480p draft is ⚡85. The same length at 720p is ⚡189, and 30 seconds at 720p is ⚡1412. New accounts get ⚡50 on signup, which is enough to try the composer and image models, not enough for a full 720p film. $10/1,000⚡ The composer quotes the exact ⚡ before every render, so nothing is charged by surprise.

Every format

One model, ten disciplines

Official ByteDance Seedance 2.5 gallery films: product, fashion, sports, stop-motion and performance. Hover to preview, press play for sound.

Product film · 9:16

The chair sells itself

Product film · 9:16

Macro that holds focus

Verbatim prompt
A richly textured series of macro close-ups focusing on an Oriental lacquer art circular installation floating on a dark, mirror-like body of water. The disk takes the form of a low, round vessel, with a warm ivory-white outer rim and a deep blue-green lacquered interior. Its surface is adorned with powdery lavender hydrangea petals, tiny wildflowers, moss, and a few scattered water droplets, resembling a miniature Eastern floral landscape drifting upon the water.
Fashion film

Three dancers, strict grayscale

Fashion film

Twelve models, one grid

Sports film

Real physics on the wall

Stop-motion style

Felt, fibers and frame steps

Verbatim prompt
Generate a 10-second warm and healing stop-motion-style video of small animals. The scene is a handmade miniature grassland set, with soft morning sunlight falling on a small meadow full of wildflowers. The little rabbit and squirrel in the frame have an adorable handmade felt / wool-felt toy texture, fluffy yet with clearly visible handcrafted fibers, and their movements show a subtle frame-by-frame stop-motion feel.
Montage

Five decades in one room

Verbatim prompt
Generate a high-tempo fast-cut short video themed around the changes in interior decoration styles of ordinary Chinese residential apartments since the founding of New China. 16:9, strong rhythm, clear visual impact, and smooth transitions between eras, like an “era transformation montage” short film. Centered on the same living room, show the evolution of residential aesthetics across different decades through fast transitions, instant furniture replacements, changes in wall colors, lighting changes, and the appearance and disappearance of home appliances.
Text to Video · performance

A performance, not a loop

Text to Video · 9:16

Camera moves, named on frame

Text to Video · voice-over

Six shot sizes in twelve seconds

Three steps

Render a take in one sitting.

01

Write the beats

Describe the shot order, the camera and the sound. Set the length.

02

Attach references

Add stills, clips or audio tracks. The model holds them across the take.

03

Render

Pick 480p or 720p. Confirm the ⚡ quote. Render.

Frequently asked questions

Open the composer, pick Seedance 2.5, and write your prompt. Set the length anywhere from 4 to 30 seconds, choose 480p or 720p, and pick 16:9, 9:16 or 1:1. Attach your reference stills, clips or audio. The composer shows the exact ⚡ price, then you confirm. Renders land in ~3–4 min and save to your library.

Seedance 2.5 is free to try, and video renders are paid per clip. A new account gets ⚡50 on signup with no card. A Seedance 2.5 take costs more than that grant, so top up when you want the full film. $10/1,000⚡ Image models and short drafts stretch the grant furthest.

No. There's no public checkpoint to download and run on your own machine. Seedance 2.5 is a hosted model, so you reach it through a platform that licenses it. OpenClips is one route: sign in, pick the model in the composer, and render in the browser. Your clips land in one library beside every other model you use.

Renders start at ⚡85 for a 4-second 480p draft. The same length at 720p is ⚡189. A full 30-second take is ⚡636 at 480p and ⚡1412 at 720p. Every step in between has its own price, and the composer shows the exact ⚡ before you confirm. Draft at 480p, then spend on the 720p pass.

Up to 30 seconds in a single take, with 4 seconds as the floor. The composer offers quick picks at 5, 10 and 15 seconds, and you can dial anything in between. ByteDance extends single-pass generation from 15 to 30 seconds. Multi-round extension appends shots to existing outputs, so a story runs past one take.

Yes. Audio is native, made in the same pass as the picture rather than dubbed on later. Dialogue, footsteps and room tone arrive locked to the frames they belong to. You can also attach audio references in the composer, and the model holds them across the take. No separate sound pass, no sync fix afterwards.

Pick 2.5 for length and references, 2.0 for resolution. Seedance 2.5 runs 4–30s in one take at 480p or 720p. Seedance 2.0 caps at 4–15s and renders up to 4K, which matters when the deliverable is a finished spot. Both carry native audio. Drafting on 2.5 and finishing on 2.0 works, since one wallet covers both.

One take. 30 seconds. Sound included.

Write the beats. Attach your references. Render at 720p.

Generate with Seedance 2.5

Free account. ⚡50 on signup. No card.

One take instead of a timeline

You write the beats. Seedance 2.5 renders them as one continuous take, 4 to 30 seconds long, with audio made in the same pass. Nothing gets stitched. The cast, the room and the light stay put from first frame to last, because there is no cut to reconcile.

That's the jump from Seedance 2.0. ByteDance extends single-pass generation from 15 to 30 seconds.

References do the art direction

Drop in the look you already have. Feed up to 30 images in one pass. Plus 10 video clips and 10 audio clips. A still pins a face or a package. A clip donates a camera move. An audio bed sets the rhythm. Name each one in the prompt and it holds for the whole take.

Timestamps, not re-rolls

ByteDance added precise editing via timestamps. Write your segments with the action, the framing and the sound for each. After the render, change one stretch and leave the rest untouched. That's the difference between a revision and a fresh roll of the dice.

What OpenClips renders

480p for drafts, 720p for delivery, in 16:9, 9:16 and 1:1. A 4-second 480p draft is ⚡85. Ask for the full 30 seconds at 720p and the composer quotes ⚡1412 before you confirm. When the deliverable needs 4K, that's Seedance 2.0 territory. When you want 1080p fast, use Seedance 1.5 Pro. For silent motion tests on the cheap, there's Seedance 1.0 Pro.

Where to go next

Compare every version on the Seedance family page. Mix vendors in one composer at the AI Video Generator. One wallet, one library, every model.