Create videos up to 30 seconds long with text, image, video, and audio references. Add up to 50 reference assets, write prompts in more than 10 languages, and make targeted edits without rebuilding the entire video.
Output Preview
Browse example prompts and video outputs to see how image, video, and audio references can guide scenes, motion, pacing, and visual direction.
Seedance 2.5 is an AI video generation model from ByteDance's Doubao team for creating longer videos from prompts, images, video clips, and audio references. It supports up to 30-second single-generation videos, targeted editing, video continuation, and multilingual prompting.
Create longer videos, work with more references, revise selected details, and describe your ideas in your preferred language.
Create a single video up to 30 seconds long. Longer output gives you more room for actions, emotional changes, and complete story beats while reducing manual clip assembly.
Combine up to 30 images, 10 video clips, and 10 audio files to guide characters, scenes, motion, pacing, and sound.
Change selected backgrounds, products, characters, or visual details while keeping the surrounding composition and timing as consistent as possible.
Describe scenes and camera directions in Chinese, English, Spanish, Indonesian, Malay, Thai, Arabic, Portuguese, Vietnamese, Japanese, Korean, and other supported languages.
Create a video in three steps: write a prompt, add reference files, and choose your generation settings.
Describe the subject, action, setting, camera movement, visual style, and sound. Include the details you want the model to preserve.
Upload supported images, videos, or audio files. Type @ in the prompt to connect each reference to a subject, scene, movement, or sound.
Choose the available settings, create the video, and use targeted editing to revise selected moments when supported by the workflow.
Use prompts and multimodal references to create longer scenes, guide visual direction, extend footage, and revise selected details.
Extend shorter footage into a longer sequence while using the original characters, scenes, lighting, and camera direction as continuity references.
Add a music track, voice recording, or sound effect without an image or video reference. The model can use audio to guide pacing, beat synchronization, physical actions, and lip movements.
Reuse clear references to guide character appearance, clothing, products, environments, and visual style across related scenes. Exact matches are not guaranteed.
Adjust selected backgrounds, characters, products, or colors without regenerating the entire video. Refine visual details while preserving the broader scene structure.
Use prompts and reference files to develop narrative scenes, educational content, advertising demos, and product concepts.
Develop short narratives, social clips, brand stories, and visual concepts with up to 30 seconds of continuous generation.
Turn product images, brand assets, scripts, and audio references into advertising concepts, campaign hooks, and product demonstrations.
Visualize products, interface flows, packaging, and physical details before committing to a full production shoot.
Create visual explanations, instructional scenes, onboarding concepts, and training demonstrations from scripts and reference materials.
Create longer sequences, work with more references, and revise selected details with fewer manual production steps.
Start CreatingGenerate a longer continuous sequence and reduce the number of short clips that need to be created and joined manually.
Reuse the same references to guide character appearance, clothing, lighting, products, and camera direction across related scenes.
Test storyboards, product shots, campaign concepts, and camera directions before arranging a full shoot or animation workflow.
Target selected details instead of rewriting the full prompt. Depending on the change, this may reduce repeated generations and credit use.
Compare published generation limits, reference capacity, extension workflows, and editing capabilities.
| Capability | Seedance 2.0 | Seedance 2.5 |
|---|---|---|
| Single-generation duration | Up to 15 seconds | Up to 30 seconds |
| Published image references | Up to 9 images | Up to 30 images |
| Published video references | Up to 3 videos | Up to 10 videos |
| Published audio references | Up to 3 audio files | Up to 10 audio files |
| Audio-only reference | Not supported as a standalone reference | Standalone audio reference supported for pacing, beats, actions, and lip movements |
| Video extension | Standard generation workflow | Temporal extension for longer sequences |
| Editing focus | Reference-guided generation | Targeted editing of selected visual elements |
| Native prompt languages | Standard language support | More than 10 supported languages |
Published capabilities do not guarantee the same result for every prompt. Test both models with your own reference materials and workflow requirements.
Seedance 2.5 is a professional multimodal video creation model developed by ByteDance’s Doubao team. It supports single-generation videos up to 30 seconds, up to 50 multimodal references, targeted editing, video extension, and native prompting in more than 10 languages.
Write a prompt, upload supported reference files, choose the available duration, resolution, and aspect ratio, then select Generate Video.
New users receive free trial credits upon registration, enough to generate one 5-second video. View the pricing page for current credit packages and generation costs.
Seedance 2.5 increases single-generation duration from 15 to 30 seconds, expands multimodal reference capacity, adds audio-only reference workflows, and provides stronger targeted editing and extension capabilities.
Seedance 2.5 supports written prompts and image, video, and audio references. You can combine supported media types in one project, subject to the current platform limits.
Seedance 2.5 supports native prompting in more than 10 languages, including Chinese, English, Spanish, Indonesian, Malay, Thai, Arabic, Portuguese, Vietnamese, Japanese, and Korean.
Seedance 2.5 supports single-generation videos up to 30 seconds long. It also supports temporal extension workflows for longer sequences.
Commercial use depends on the current platform terms, your rights to all uploaded materials, and applicable law. Review the Terms of Service before using generated content in advertising, client work, or paid products.
Real human faces cannot be used as multimodal references on this platform. Use virtual characters or AI-generated faces instead.
Create cinematic time-freeze videos with Seedance 2.5 using five practical prompts, POV, car-chase, crowd, rewind workflows, and quick fixes.
Compare Seedance 2.5 and Wan 3.0 through real tests of 30-second storytelling, document-to-video, references, physics, acting, action, cost, and use cases.
Explore how Seedance 2.5 supports AI advertising, product demos, game previsualization, training videos, SOP content, targeted editing, and synthetic scenarios.
Add a prompt and your reference files, choose the available generation settings, and create a video up to 30 seconds long.