Longer ideas
Develop a complete sequence instead of assembling disconnected short clips.
SEEDANCE 2.5
AI video generation built for complete scenes
Create expressive video up to 30 seconds with picture and sound generated together.MODEL OVERVIEW
What is Seedance 2.5?
Seedance 2.5 is ByteDance's next-generation AI video model for creating longer, more coherent scenes from prompts and multimodal references.
It combines multi-shot storytelling, stronger subject continuity, and native audio generation in one workflow—giving creators more control from concept to output.
Stronger direction
Guide character, style, framing, motion, and atmosphere with reference assets.
Picture and sound
Generate visuals, dialogue, ambience, and effects as one coordinated result.
01 · PRECISE REFERENCE CONTROL
Turn references into creative direction
Combine up to 50 reference assets—including images, video, and audio—to communicate the intended character, composition, visual style, movement, and atmosphere. Use references to keep products recognizable, characters consistent, and each shot closer to the original creative direction.02 · NATIVE AUDIO AND PERFORMANCE
Create picture and sound as one experience
Generate visuals and audio together instead of treating sound as a separate step. Dialogue, ambience, effects, and performance can develop alongside the image. Create narrative content, social campaigns, and product stories where sound is part of the idea from the beginning.03 · SCENE TRANSFORMATION
Carry the same characters into a new world
Move a scene from a sunlit park to a vineyard while preserving the characters, their relationship, and the natural rhythm of their movement. Use reference-led transformation to explore new settings without losing the identity and visual continuity that connect the sequence.HOW TO USE SEEDANCE 2.5
From creative brief to generated video in three steps
Describe the scene
Write the subject, action, camera direction, visual style, and desired sound.
Add references
Provide images, video, or audio to guide characters, products, motion, and tone.
Generate and refine
Create the video, review the result, and adjust the prompt or references as needed.
PRIVATE ASSET WORKFLOWS
Generate with verified people or reusable virtual characters
Consent and identity verification before generation
Create a private verified group, register approved reference media and use the resulting asset only after it reaches Active status. Available to approved tenants.
Persistent fictional characters for repeatable content
Reuse private virtual character assets to improve continuity across series, original IP and branded content workflows.
BUILT FOR DEVELOPERS
Bring Seedance 2.5 into your product through one API
Simple asynchronous workflow Up to 50 multimodal references 480p, 720p and 1080p direct output Usage-based pricing
FREQUENTLY ASKED QUESTIONS
Seedance 2.5, explained
What is Seedance 2.5?
Seedance 2.5 is ByteDance's AI video generation model for longer multi-shot scenes, multimodal reference control, stronger continuity, and native audio.
What inputs can I use?
You can create from text and guide the result with image, video, and audio references. The LSF integration supports up to 50 reference assets per request.
How long and what resolution can the output be?
Seedance 2.5 can generate video up to 30 seconds. Native output supports 480p, 720p and 1080p under the enabled native model contract. Use LSF Video Enhancement when you need a separate enhancement workflow for an existing video.
Does Seedance 2.5 generate audio?
Yes. It can generate picture and synchronized sound together, including dialogue, ambience, and effects guided by your prompt and references.
Can I use private real-human or virtual assets?
Yes, when the capability is enabled for your tenant. Real-human assets require the applicable consent, identity verification, media registration, review and Active status before generation.
Which model name should my integration use?
Use the exact tenant-specific model alias provided by Light Speed Future during onboarding. Do not infer an internal provider model or mapping.
BRING YOUR IDEAS TO LIFE