How to use Seedance 2.5: your first AI video
Use Seedance 2.5 to turn a written scene or a reference image into a short video. This guide walks through the controls available in SWRFlow so you can prepare one clear shot, check its estimated cost and review the result before making another version.
The starting point
Start with one subject, one visible action and one camera move. Choose Seedance 2.5 in the video generator, set a short duration, review the credit estimate and generate. Keep the prompt and settings with your result so the next revision has a clear starting point.
On this page
Before you begin
- A SWRFlow account and enough credits for the selected settings.
- A short scene description, or an image of an object, landscape or illustrated character.
- A specific goal for the first shot: what should move, and what should stay the same.
1. Choose text or a reference image
Open the AI video generator and explicitly select Seedance 2.5. The model currently selected on the page may be different. Use text-to-video when the scene does not yet exist. Choose image-to-video when you already have a composition or subject you want the model to use.
| Your starting point | Use | What to describe |
|---|---|---|
| An idea without artwork | Text-to-video | Subject, setting, action, camera and optional sound. |
| A product image or illustration | Image-to-video | The intended motion and the details to preserve. |
A reference is guidance, not a guarantee that every pixel will remain unchanged. For a first attempt, use one clean image and avoid small lettering, overlapping objects or several competing subjects. Do not assume that an uploaded image is a precise first-frame control in every model or mode.
2. Write a Seedance 2.5 prompt around an action
Write what happens, rather than a long list of quality adjectives. If the source image already establishes the scene, use most of the prompt to describe movement. Keep camera movement separate from subject movement so you can change one without changing the other.
Use the reference image as the visual guide. A single blue paper boat drifts slowly across a shallow pool. Small ripples spread behind it. The camera stays at water level and follows gently from the side. Preserve the boat's folded shape, blue color and proportions. Soft daylight, a quiet atmosphere, one continuous shot.
How to adapt itStarting template: use your own paper-boat image. Try a short clip first. If the shape changes, simplify the motion and reduce the camera movement.
Read the prompt once as a shot instruction: is there one visible event, and can it reasonably happen in the selected duration? Remove anything that asks the subject to perform several unrelated actions at once.
3. Set duration, shape and resolution
In the current SWRFlow integration, Seedance 2.5 offers 4–30-second clips at 480p, 720p or 1080p. These are the options exposed by this platform; availability in other products can differ. Choose the output aspect ratio for where the video will be used, such as 16:9 for a wide frame or 9:16 for a vertical composition.
- Begin with 5 seconds when the action is simple. Use a longer clip only when the shot needs time to develop.
- Use a lower available resolution to explore the idea, then review the estimate again before requesting a larger output.
- Check the image-to-video mode and references after changing models. Different models expose different inputs and limits.
- Use the controls for duration and resolution. Writing a different setting in the prompt does not change the selected request.
The quick generator accepts image references. SWRFlow currently allows up to 16 for this model, but one useful reference is a better starting point than a full upload limit. Audio references, a dedicated first/last-frame mode and video extension are not available in this SWRFlow workflow. Real-person photo references also need a review workflow that is not currently offered here; use objects, landscapes or illustrations.
4. Review credits and generate
Look at the credit estimate after you finish selecting the model, duration and resolution. Increasing duration or quality can change the cost substantially. Make sure the account has enough credits before submitting. The generator's current estimate is the amount to check, rather than a number copied from a tutorial or another platform.
Once submitted, let the current task reach a clear outcome. If processing appears slow, inspect the task status before pressing Generate again: another submission may create another paid task. A successful generation that does not match your creative taste is different from a failed generation; consult the site's current credit and refund information for how each is handled.
5. Review, download and make one revision
- Watch the full clip at normal speed. Check whether the main action occurs and whether the camera behaves as requested.
- Pause near the beginning, middle and end. Look at the subject's silhouette, object count and any small details that matter.
- Listen if the output has audio. Do not judge the visual result alone when sound is part of the deliverable.
- Download the completed result and save the prompt, model and settings beside it.
- Choose one issue for the next revision. For example, keep the same image and prompt but ask for a stationary camera.
If you need more variations, the prompt guide gives separate templates for object movement, camera direction and atmosphere. For a product-led shot, continue with the product-image-to-video workflow and define which packaging details must survive.
Questions
Can I generate a 30-second video with Seedance 2.5?
SWRFlow currently exposes durations from 4 to 30 seconds for Seedance 2.5. Select the duration in the generator and check its estimate. Start shorter when you are testing an uncertain idea; a longer clip also gives the model more actions and continuity to manage.
Can I extend an existing video in this workflow?
This guide covers the Seedance 2.5 modes currently available in SWRFlow. Dedicated video extension and editing are not exposed here. Capabilities described for the underlying model or another provider are not a promise that the same controls exist in this generator.
Does a longer prompt give a better result?
Length alone is not the goal. Add detail when it resolves a real decision, such as which object moves or what the camera follows. Contradictory instructions and too many events can make a shot harder to interpret. Begin with a clear event and refine one variable at a time.
References & useful links
- ByteDance: introducing Seedance 2.5
Background on the underlying model. SWRFlow exposes the subset of controls described in this guide.
Put the next step into practice
Select Seedance 2.5, choose a short duration and check the estimate before generating.
Open the video generator