Now available on LinkModel

Seedance 2.5 API

Thirty seconds. One continuous story. Native sound included.

ByteDance's multimodal video model turns text, images, video, and audio references into controlled 4–30 second sequences with native sound.

Try Seedance 2.5View API docs

30 sec

Maximum standard duration

480p / 720p

LinkModel listed output

30 + 10 + 10

Image, video, and audio references

A/V

Native picture and sound

Price comparison

Run Seedance 2.5 for less on LinkModel

Save up to 54.6% compared with fal.ai. Compare 5-second, 16:9 pricing across providers.

Seedance 2.5 pricing by provider
Generation spec
fal.aiCompetitor
AtlasCloudCompetitor
WaveSpeedCompetitor
LinkModelOur price
480p16:9
$1.0747
$0.670
$0.90
$0.4883
Save54.6%vs fal.ai
Lowest price
720p16:9
$2.3112
$1.5028
$1.80
$1.0982
Save52.5%vs fal.ai
Lowest price

Start saving with Seedance 2.5

Try Seedance 2.5View API docs

Creation modes

Three ways to create

Start from a written scene, an existing visual, or a multimodal reference stack. The live endpoint determines which inputs and controls are available for each workflow.

ModeInputBest forControl
01Text to VideoText promptCreate a scene from scratchScene, action, camera, and sound
02Image to VideoOne or more imagesAnimate an existing visualSubject, composition, and art direction
03Reference to VideoImages, video, and audioKeep assets consistent across shotsIdentity, movement, style, and sound

Timeline proof

Thirty seconds, directed beat by beat

Treat a generation like an edit: establish context, develop movement, stage the turn, and land the final frame.

A barber shaping an elderly client's haircut in a barbershop mirror

Direction canvas / up to 00:30

Beat 01 / 04

0s

Establish

Production controls

More room to direct, fewer resets

Compose references, sound, events, and continuity as parts of one production brief.

Reference stack

Multimodal reference control

Use up to 30 images, 10 videos, and 10 audio clips to shape subject, style, motion, and sound.

Audio direction

Audio-driven creation

Let audio participate in the creative direction, including workflows driven by audio alone.

Timed beats

Timestamp direction

Place actions and changes at explicit moments across a longer sequence.

Continuity

Longer narrative continuity

Build a complete 4–30 second story unit with more room for setup, development, and resolution.

Use cases

See longer narratives, reference control, and precise edits

Watch official Seedance 2.5 demonstrations alongside a LinkModel-generated product narrative.

A dreamlike forest opening into a luminous underwater world

30-second narratives

Official demonstration footage follows one continuous result from a train window through a blooming forest, an underwater world, and a quiet interior.

An ensemble cast gathered in an elegant ballroom

Multimodal reference scenes

Official demonstration footage coordinates an ensemble cast from multiple character references while preserving roles, wardrobe, and blocking.

A beauty creator demonstrating lipstick in a bedroom studio

Precise video edits

Official demonstration footage replaces a presenter while preserving the surrounding shot, speech, product action, and camera framing.

API quickstart

One API, an asynchronous workflow

Submit a generation request, poll the task until it completes, then retrieve the generated video from the result. The same LinkModel video API keeps that lifecycle consistent across supported models.

View API docs

Quickstart

Make your first request

curl -X POST https://api.linkmodel.ai/v1/videos/generations \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2-5",
    "prompt": "A continuous cinematic sequence with synchronized ambient audio"
  }'

Model family

Explore the Seedance family

Seedance 2.5 is the primary destination. Other family versions remain visible as catalog context, with actions shown only when that exact model is available.

Featured model

seedance-2-5

Latest

The primary Seedance workflow for longer multimodal sequences with native picture and sound.

30 sec
Maximum standard duration
480p / 720p
LinkModel listed output
30 + 10 + 10
Image, video, and audio references
A/V
Native picture and sound
Try Seedance 2.5View API docs

Seedance family

Family catalog

03

Related guides

Go deeper with Seedance 2.5

Compare versions, plan a 30-second prompt, estimate cost, and move from evaluation to API implementation.

FAQ

Frequently asked questions