Seedance 2

byByteDance

Continuous 30-second single shots with synchronized audio and multi-reference control

Direct video
Seedance 2

How Seedance 2 works

From raw multi-asset references to continuous 30-second takes with native multichannel sound.

Script prompt and media tags

Script prompt and media tags

Draft your scene description with timestamp intervals and reference tags like @Image1, @Video1, or @Audio1 for identity, motion, and voice control.

Select model tier and duration

Select model tier and duration

Choose Seedance 2.5 for up to 30 seconds of audiovisual generation, or pick 2.0 Fast or Mini for rapid storyboard previsualization up to 15 seconds.

Configure framing and generate

Configure framing and generate

Set your aspect ratio, resolution from 480p to 4K, and bitrate mode, then render the continuous video with synchronized audio.

What Seedance 2 is good at

Built on unified diffusion transformer architecture for coherent long-form video, identity retention, and acoustic synchrony.

Unified 30-Second Single Shots

Unified 30-Second Single Shots

Generate continuous takes up to 30 seconds on Seedance 2.5 without iterative stitching, preserving lighting consistency, spatial depth, and physical momentum.

Multimodal Reference Control

Multimodal Reference Control

Lock character identity across up to 30 reference images and transfer motion from up to 10 video clips using explicit @Image, @Video, and @Audio prompt tags.

Synchronized Multichannel Audio

Synchronized Multichannel Audio

Co-generate native dialogue, ambient environment tone, and action-timed foley effects in a single pass, or feed reference audio tracks to steer lip-sync timing.

Fast and Mini Rapid Iteration

Fast and Mini Rapid Iteration

Switch to 2.0 Fast or Mini tiers for low-compute storyboard previsualization and motion validation up to 15 seconds before committing final takes to the flagship 2.5 engine.

Made with Seedance 2

Explore continuous single-shot cinematography, multi-reference character consistency, and synchronized sound design produced across Seedance engines.

High-speed beverage commercial with synchronized liquid acoustics

High-speed beverage commercial with synchronized liquid acoustics

Atmospheric 30-second performance shot with continuous orbit choreography

Atmospheric 30-second performance shot with continuous orbit choreography

Architectural lighting study showing volumetric light transitions

Architectural lighting study showing volumetric light transitions

Vertical short-form athletic choreography with synchronized foley

Vertical short-form athletic choreography with synchronized foley

Editorial wildlife narrative with crisp aerodynamic sound design

Editorial wildlife narrative with crisp aerodynamic sound design

What people build with Seedance 2

How directors, fashion houses, and product designers harness continuous shot generation and rapid previsualization.

Narrative Filmmaking and Scene Sequences

01

Direct continuous 30-second takes with complex multi-action blocking, fluid crane and dolly paths, and native dialogue co-generation without multi-clip assembly.

Fashion Lookbooks and Character Campaigns

02

Lock wardrobe styling and model likeness across multiple angles using up to 30 reference stills while applying fluid runway motion from video references.

Automotive and Industrial Product Spots

03

Showcase mechanical hardware and vehicle dynamics with realistic inertia, surface reflections, and synchronized mechanical audio from 480p up to 4K resolution.

Commercial Previsualization and Storyboarding

04

Iterate director treatments and shot pacing rapidly using 2.0 Fast and Mini engines before rendering high-bitrate hero deliverables on Seedance 2.5.

Social and Short-Form Vertical Content

05

Produce native 9:16 vertical motion clips with punchy foley and beat-matched rhythm for mobile campaigns and digital storytelling.

Frequently Asked Questions

Seedance 2.5 generates continuous single-shot video up to 30 seconds with native co-generated multichannel audio, audio reference input, and support for up to 30 reference images and 10 reference videos. Seedance 2.0 caps at 15 seconds duration, accepts up to 9 reference images and 3 reference videos, does not support audio reference ingestion, and offers dedicated Fast and Mini tiers for budget-conscious iteration.

Use Seedance 2.0 Fast or Mini during early storyboarding, layout testing, and prompt syntax validation where turnaround speed and compute efficiency matter more than micro-detail. Once choreography, lighting cues, and reference mappings are locked, switch to standard Seedance 2.0 or flagship Seedance 2.5 for maximum visual fidelity and rich acoustic layers.

Seedance uses explicit reference syntax in the text prompt to direct specific media inputs. By referencing @Image1, @Video1, or @Audio1, you can independently assign character facial identity and costume styling from still images, camera paths and choreography from video clips, and vocal performance timing from audio tracks.

No, audio-only reference generation is not supported. When uploading reference audio files to Seedance 2.5, you must also provide at least one reference image or video asset so the model can visually map the performance, lip-sync, and environment.

Seedance struggles with fine typography on distant background signs, rapid multi-hand intricate finger interactions, and pure 2D graphic vector animations. Additionally, extreme 30-second continuous shots may occasionally exhibit subtle micro-feature drift across complex multi-subject interactions.

Try Seedance 2 on Fuser

Continuous 30-second single shots with synchronized audio and multi-reference control

Direct video