Announced 30-second output
The Seedance 2.5 preview targets a complete cinematic beat in one continuous generation, with more room for story and motion.
Seedance 2.5 model preview
Seedance 2.5 PreviewAI Video Model Guide
Explore the announced Seedance 2.5 capabilities, including up to 30 seconds of continuous generation, 50 multimodal references, and precise local video editing.
30-second generation50 referencesLocal editing
Preview information: final specifications and Filtrix availability may change as the model rolls out.
Create with Filtrix
Explore the current AI video workflow
A cinematic tracking shot through a rain-soaked city at night...
Duration
Announced: up to 30s
Model
Seedance 2.5 preview
Key features
What makes Seedance 2.5 different
More time, more reference control, and a more practical way to revise generated video.
Up to 50 references
Public preview information describes richer multimodal guidance for characters, products, locations, style, camera movement, and sound.
Consistent local editing
Refine a selected detail while preserving the composition and continuity of the rest of your video.
Stronger consistency
Keep character identity, wardrobe, setting, and visual direction stable throughout longer scenes.
Cinematic camera control
Describe shot size, camera movement, pacing, and transitions in natural-language prompts.
Multimodal direction
Bring visual and audio references together as one clear creative brief for the model.
Designed for creators
More control at every stage of the shot
Long-form generation
Create a full scene, not just a short moment
Seedance 2.5 was previewed with single-generation video up to 30 seconds. That extra time could give actions space to develop, let camera moves feel intentional, and make it easier to build a beginning, middle, and end without manually stitching multiple clips.
Continuous 30-second scenesLonger camera movementMore complete storytelling
Multimodal references
Give the model a real creative brief
The announced workflow supports up to 50 reference assets to communicate what words alone cannot. That could include character images, product angles, environments, visual styles, motion examples, and sound references.
Characters and productsStyle and locationsMotion and audio direction
Local video editing
Keep the take you like and fix what you do not
The previewed local-editing workflow is intended to revise a selected area while maintaining consistency across surrounding frames, instead of regenerating an entire scene for one unwanted detail.
Targeted revisionsPreserved compositionFaster creative iteration
Use cases
Made for modern video creation
Short films
Create longer narrative scenes with characters, camera movement, and a clear dramatic beat.
Advertising
Keep products, talent, brand colors, and visual style aligned across a polished commercial shot.
Social content
Turn a concept or key visual into engaging vertical and landscape video for social campaigns.
Creative concepts
Visualize treatments, storyboards, and campaign ideas before committing to full production.
How it works
Create a video in three simple steps
Upload or describe
Start from an image or describe the scene you want to create with a detailed text prompt.
Add creative direction
Define the action, camera movement, style, pacing, and references that should guide the result.
Generate and refine
Create your video, review the result, and adjust the prompt or selected details as needed.
FAQ
Frequently asked questions
Seedance 2.5 is an announced update to ByteDance's AI video model family. Public preview information highlights continuous videos up to 30 seconds, support for up to 50 multimodal references, and consistency-preserving local video editing. Final specifications may change at release.
Bring your next video idea to life
Create polished AI video from an image and a prompt with the simple Filtrix creative workflow.