Seedance 2.5 is now officially released
From short clips to complete stories with Seedance 2.5
Seedance 2.5 advances unified multimodal audio-video generation from isolated clips toward complete creation. It can produce a 30-second high-quality clip in one pass, extend a result over multiple rounds, organize connected shots into a coherent story, and improve image, sound, motion, transitions, and material realism. Its expanded reference and editing tools give creators more control over complex subjects, camera movement, pacing, and revisions. This announced model is not yet available here; the live workspace uses Seedance 2.0 Standard and Seedance 2.0 Fast.
What the latest release brings to video creation
Use Seedance 2.5 when a concept needs more than a short animated moment. The model is built for longer narrative progression, multimodal art direction, multi-subject consistency, timestamp-level pacing, and targeted edits after generation. Official examples span one-take performances, ensemble scenes, green-screen replacement, education, and industrial visualization.
Seedance 2.0 vs Seedance 2.5
The generational change is about moving from controlled clip creation to a more complete production workflow. Seedance 2.5 keeps the unified multimodal foundation while substantially expanding duration, reference capacity, narrative continuity, and editing precision.
Seedance 2.0
Seedance 2.0 is built for quick generation and repeated drafts. It creates 4–15-second videos with Standard and Fast modes, first- and last-frame guidance, and references of up to 9 images, 3 videos, and 3 audio clips.
Seedance 2.5
Seedance 2.5 doubles single-pass duration to 30 seconds, supports multi-round extension, accepts up to 30 images, 10 videos, and 10 audio clips, and adds timestamp-level control plus targeted character, action, sound, story, green-screen, perspective, and camera edits.
Ready to create now? You can try the Seedance 2.0 multimodal workflow today and explore the newest Seedance 2.5 capabilities through the official examples on this page. We will update the online experience as soon as Seedance 2.5 becomes available.
Generation modes
Seedance 2.5 supports foundational generation and reference-led creation. Start from text, guide a shot with an opening or ending frame, or combine images, video, and audio when identity, motion, rhythm, sound, and camera language must work together. Official limits reach 30 images, 10 videos, and 10 audio clips in one reference request.
Text to video
Describe the subject, environment, action, camera movement, lighting, and mood without uploading media. Text-to-video is useful for concept exploration, establishing shots, visual storyboards, and ideas that do not need to preserve an existing product or character image.
Multimodal reference to video
Guide Seedance 2.5 with up to 30 images, 10 video clips, and 10 audio clips in one request. The model can combine composition, setting, style, characters, props, voices, motion, rhythm, and sound to direct complex multi-subject and multi-scene work instead of treating one image as the only visual anchor.
First-frame image to video
Upload one image as the opening frame and describe what should happen next. The model uses that frame as the starting composition, making this mode suitable for animating a product photo, portrait, illustration, landscape, or designed campaign visual.
First and last frame to video
Provide an opening image and an ending image to define the visual transition. The prompt should explain the motion between them, helping the model create a shot with a clearer destination than a first-frame-only generation.
More official creations
The 30-second one-take concert now appears in the generation preview. Continue with the complete creative film, multi-round extensions, multimodal references, green-screen changes, camera edits, and education and industry examples to see where Seedance 2.5 fits your next project.
Typical workflow
A strong Seedance 2.5 workflow treats the prompt as a timed production brief. Define the references and story beats, assign actions and camera moves to time ranges, generate up to 30 seconds, then extend or edit only the section that needs refinement.
- 1Upload or choose a reference image to anchor the visual style.
- 2Write a motion prompt describing camera movement, subject action, mood, and pacing.
- 3Select the speed, duration, resolution, aspect ratio, and audio options.
- 4Generate a short video draft, review it, and reuse the prompt for the next variation.
How to write a prompt that holds the shot
A useful prompt describes visible change instead of relying on broad adjectives. Write the shot in a logical order: identify the subject, specify its action, define camera movement, describe the environment and lighting, then state the pacing or final composition you want.
Anchor the subject
Name the main subject and the details that must remain recognizable, such as product shape, clothing, facial direction, logo placement, or illustration style. Clear visual anchors reduce ambiguity when the model adds motion to a still image.
Separate subject and camera motion
Explain what the subject does and how the camera responds. For example, describe a bottle rotating slowly while the camera pushes in, or a person turning toward the light while the shot tracks from left to right.
Control pacing and continuity
Use timing words such as slowly, steadily, in one continuous take, or with a gentle finish. When using first and last frames, describe the transition between them so the model understands the intended path instead of inventing unrelated changes.
Write for the destination format
Mention whether the result is a vertical social ad, a widescreen product reveal, a quiet brand loop, or a fast concept draft. The intended use helps align framing, movement intensity, empty space, and overall visual rhythm.
Choose the right settings before you create
Choose Fast when you want to test prompts and formats quickly, or Standard when resolution and final visual quality matter more. Clips from 4 to 15 seconds keep motion and pacing easy to direct, while longer options give an action more room to develop. Turn on audio or web assistance only when it adds something useful to the idea.
See credit pricingUse cases
Seedance 2.5 fits projects that need a complete narrative beat, many coordinated references, or precise changes without rebuilding the entire video. It can support campaign storytelling, multi-character performances, educational visualization, synthetic industrial footage, and directed edits for film or advertising workflows.
For commerce teams
Product videos for e-commerce pages, launch teasers, and paid social ads.
For creators and agencies
Portrait, lifestyle, and UGC-style clips that need natural subject motion.
For creative testing
Rapid creative testing when a team needs several image-to-video directions.
Frequently asked questions
Use these quick answers to decide whether Seedance 2.5 fits the video you want to make.
What is new in Seedance 2.5?
Seedance 2.5 adds 30-second single-pass generation, multi-round extension, stronger long-form narrative continuity, much larger multimodal reference sets, timestamp-level control, and targeted post-generation editing.
How many references can Seedance 2.5 use?
According to ByteDance's official release, one reference request can include up to 30 images, 10 video clips, and 10 audio clips. The model combines those materials to guide subjects, settings, style, voices, motion, and camera direction.
Can I try Seedance 2.5 now?
You can watch the complete official examples now to explore 30-second storytelling, multimodal references, and targeted editing. Online generation currently uses the available Seedance models, and we will update the experience when Seedance 2.5 becomes available.