What will you create next?
Plan a longer scene with Seedance 2.5 using text and reference media. Choose output up to 4K and give the action a clear beginning, development and ending.
Seedance 2.5 · Updated August 2026
Seedance 2.5 supports scenes up to 30 seconds with output up to 4K and a larger reference set. Plan the beginning, main action and ending before choosing the materials that define the subject, movement and sound.
30s
single short-film generation
4K
native resolution target
50
full-modal reference inputs
A single clip can run for 30 seconds, giving an action time to develop from its opening composition to a deliberate final frame.
Choose output up to 4K when a scene needs more room for close-up detail or a larger final presentation. Inspect the generated textures and edges at full size.
The workspace supports 30 images, 10 videos and 10 audio files. Assign each reference a role in the brief, such as identity, location, camera motion or sound.
Use film, television or character references only within the rights and permissions you hold. Model access does not itself grant a license to someone else's characters or material.
Describe the scene spatially: where subjects stand, how they move and the path the camera follows. This can support pre-visualization and blocking before a finished production.
Use real product references to explore a reveal, demonstration or launch scene. Check packaging, material and identifying details before using the take in a campaign.
Develop a scene around characters and materials you are authorized to use. Keep the intended action and reference roles clear, and confirm the necessary rights for the project.
Plan a 30-second narrative beat to test staging and camera movement. Use the result to discuss pacing and spatial relationships before committing to a full shoot or animation.
Creation workflow
A longer clip needs a readable progression. Define what changes from the first frame to the last, then choose references and settings that support that sequence.
Step 01
For a 30-second scene, give each section a purpose: 1–10 seconds establish the subject, 11–20 seconds develop the action, and 21–30 seconds resolve it. Adjust the pacing to the actual event instead of filling every second with another instruction.
Step 02
Use up to 30 images to establish subjects and settings. Add video or audio only when it contributes a specific movement, camera path, voice or rhythm to the scene.
Step 03
Choose the aspect ratio and duration, then select the required resolution up to 4K. Review the estimate for the complete request before generating.
Step 04
Inspect subject placement, camera travel, surface details and the final beat. Revise the ambiguous instruction or conflicting reference rather than adding a general request for higher quality.
Compare the models by the work they need to do. Seedance 2.5 supports longer scenes and larger reference sets; Seedance 2.0 and Mini offer shorter workflows with their own quality, resolution and credit choices.
| Capability | Seedance 2.5 | Seedance 2.0 | Seedance 2.0 Mini |
|---|---|---|---|
| Best fit | Longer product scenes, narrative studies and brand concepts that need several coordinated references. | Text and reference-led scenes in the Seedance 2.0 workspace, with quality choices for the intended take. | Short visual drafts and repeated prompt experiments where the available resolution and credit rate fit the job. |
| Reference inputs | Up to 50 references: 30 images, 10 videos and 10 audio files, each with a defined role in the scene. | Up to 9 images, 3 videos, and 3 audio files for standard multimodal generation. | Multimodal reference input for shorter scenes, within Mini's own file and output limits. |
| Duration path | Up to 30 seconds for a sequence with an opening, developed action and a resolved ending. | 4-15 second generation for stable Seedance 2.0 production workflows. | 4-15 second generation for quick iterations and mini-tier drafts. |
| Resolution path | 4K, 2K, 1K, and 720p output. | 4K, 2K, 1K, and 720p output. | 480p and 720p output for cost-efficient production. |
| Workflow role | Choose it when a scene depends on a larger set of product, character, movement and sound references. | Use it when the priority is stable production generation. | Choose it to compare shorter drafts before deciding which brief deserves further development. |
| Main advantage | Useful when longer duration, output detail and a larger reference set are requirements of the brief. | Useful for 4–15-second scenes using the established text, image and media controls. | Useful when a short scene fits Mini's resolution choices and the credit budget is a priority. |
| Specification | Seedance 2.5 |
|---|---|
| Max clip length | Up to 30 seconds in a single generation |
| Resolution | Native 4K |
| Reference inputs | Up to 50 files: 30 images, 10 videos, 10 audio tracks |
| Input modes | Text to Video, Image to Video, Media to Video |
| Audio | Native audio generated with the visuals |
| Aspect ratios | 16:9, 9:16, 1:1 |
| Commercial use | Included with paid plans |
A timeline helps a longer brief stay readable. These examples assign the opening, development and ending their own action and sound cues.
0–10s: A silver camera rests on a dark wooden table beside a folded map. Slow push-in, warm side light. 10–20s: Close views of a hand turning the focus ring and opening the map; keep the camera shape unchanged. 20–30s: Pull back as the hand lifts the camera, leaving the map in the light. Sound: quiet room ambience, a soft mechanical click and a restrained piano phrase.
0–10s: A gardener enters a glasshouse before sunrise, following shot at shoulder height. 10–20s: She opens a roof vent and waters a tray of seedlings, close views of droplets and leaves. 20–30s: The camera moves back as sunlight reaches the plants and she pauses by the door. Sound: footsteps, water and distant birds; no dialogue.
0–10s: A model in a dark blue jacket walks along a quiet riverside, vertical 9:16, soft overcast light. 10–20s: Cut to the jacket fastening, sleeve texture and a hand resting on the railing. 20–30s: The model turns toward the river as the camera settles into a still medium shot. Keep the jacket details consistent. Sound: gentle water and a sparse electronic rhythm.
Details about planning a scene, using references, choosing output settings and understanding the task estimate.
Seedance 2.5 is a video model for text and reference-led scenes. Its workspace supports clips up to 30 seconds, output up to 4K and larger sets of visual and audio references.
Draft generates a lower-resolution preview so you can assess composition, movement and camera work before paying for a higher-resolution final. Revise the prompt or references if the preview needs another attempt. A final generated from the preview keeps its content, although fine details may vary. Create the final within 24 hours of the preview. Preview and final are separate charged tasks.
A single generation can be up to 30 seconds. Give the sequence a clear opening, main action and ending so the extra time serves the scene.
4K is useful when the intended layout or edit needs more image detail. Inspect product textures, close-ups and fine edges in the generated output before selecting the final take.
The workspace limit is 50 reference materials: 30 images, 10 videos and 10 audio files. Explain which assets guide identity, setting, motion or sound instead of leaving their roles implicit.
Commercial use is included with annual plans, monthly Standard, Pro and Max, Team plans, and Creator Pack or larger packs. Free allowance, monthly Basic, Starter Pack and trial-pack outputs are for personal, non-commercial use. You also need appropriate rights for uploaded images, audio, likenesses and brand material.
Use film, television or character material only when you have the required authorization for the project. Access to the generator does not grant rights to uploaded characters, trademarks, performances or likenesses. Check the permissions attached to those materials before generating or publishing.
It can be used for product reveals, narrative studies, campaign scenes and storyboard pre-visualization. Choose it when the brief benefits from a longer clip or a larger set of references.
Its main workflow differences include clips up to 30 seconds, output up to 4K and more reference inputs. Spatial descriptions can also guide subject placement and camera paths when planning a scene.
Prepare clear subject images and only the style, motion or audio references that serve the brief. Explain their roles and keep the subject details compatible across the set.
Processing time varies with the scene, length, resolution, reference inputs and current demand. Follow the task status in the workspace; a 30-second 4K request can take longer than a shorter, lower-resolution one.
A request can combine text with up to 30 images, 10 videos and 10 audio files. Use each reference to communicate a specific part of the subject, style, movement or sound.
The Media Files bar supports local uploads, assets from your library and direct public file URLs. Upload by dragging files or selecting them from your device; use Asset to reuse existing material. URL input requires a file link such as .jpg, .mp4 or .mp3, rather than a web page or YouTube watch link.
Choose 2.5 when the brief needs more than 15 seconds, native 4K or a larger reference set. For a 4–15-second scene, compare Seedance 2.0's settings and credit estimate with the requirements of the take.
Access is available with paid plans. Consult the current pricing page for the applicable plan and credit details before submitting a task.
After a video is attached, choose its role. Reference uses it to guide a new clip; Extend continues from its ending; Edit revises the attached clip itself. Extend and Edit preserve the input aspect ratio, and Edit also follows the original length. The workspace applies those output constraints for the selected operation.
Smart chooses the ratio from the reference material. Turn it off to select one of six fixed ratios when the workflow permits. It remains enabled for image-to-video because the first frame sets the ratio, and for video Extend or Edit because the original clip sets it.
Auto duration is unknown until the task finishes, so credits are reserved against the 30-second maximum. The button shows that maximum reservation. After successful generation, the actual length determines the charge and unused reserved credits return automatically without expiring. Choose a fixed duration instead if you want a fixed estimate before submitting.
Plans start at $14.90/mo, billed annually.
Browse curated Seedance 2.0 video prompt examples with videos, authors, publish dates, and engagement metrics.