How to Use Seedance 2.5: A Five-Step Video Workflow

Getting started with Seedance 2.5 is easier when you decide what the video needs to do before you write the prompt. Choose a model and an input — text, an image or supported references — then check the duration, resolution, audio options and estimated credits before you generate.

1. Choose text-to-video or image-to-video

If you are starting with an idea, open text-to-video. If a person, product or composition needs to stay recognisable, image-to-video gives the model a visual starting point and can save you a few rounds of guesswork.

To use motion or atmosphere from a video or audio reference, choose video-to-video. Tell the model what each upload is for — for example, a reference for the subject, movement, camera or sound — so the references have a clear job.

2. Write a prompt the model can follow

A useful prompt is clear before it is long. Try this order: subject → action → setting → camera → light and mood → sound.

On a rainy evening in Bristol, a cyclist pauses beneath a shop awning. The camera follows slowly from behind as the street lights reflect on the wet pavement. A calm, natural colour palette; light rain and distant traffic.

When you review the first result, change one thing at a time. If the scene works but the camera moves too quickly, try “the camera tracks at a steady, unhurried pace”. It is easier to compare generations when you change one instruction at a time. For more examples, see the Seedance 2.5 prompt guide.

Which Seedance model should you choose?

Seedance 2.0, Seedance 2 and their search variants refer to the same model family. In this workspace, choose Seedance 2.0, Fast or Mini only when the option is shown in the model selector. The available inputs, duration, resolution, audio controls and estimated credits can vary, so use the settings displayed for your selected model.

For ByteDance's model information, visit the Seedance 2.0 project page. Seedance 2.5 UK is an independent service, not an official ByteDance or Dreamina website.

3. Check duration, resolution and aspect ratio

The generator may offer clips from 4 to 30 seconds and 480p or 720p, depending on the model and mode. Not every combination is available for every input, so check the controls shown after you choose a model.

Start with a short clip to check the motion and framing, then make a longer version once the direction feels right. Decide where the video will be posted first: vertical for a short-form feed, or landscape for a YouTube video or presentation.

4. Check the estimated credits before generating

The credit cost can vary with the model, duration, resolution, references, audio and provider route. Check the estimate beside the Generate button before submitting. Current plans and credit packs are listed on the Seedance 2.5 pricing page.

Promotional starter credits may be available to new accounts, but they are not permanent or unlimited. See our guide to free Seedance 2.5 credits for what to check before you begin.

5. Review the result and make a focused change

Assess the subject, action, camera, lighting and sound separately. Change the part that needs work, then generate again with the other settings held steady. That makes it easier to see which instruction made a difference.

If a person or product changes, choose a clearer reference and say what must stay the same. If the action feels muddled, reduce the number of things happening in one shot. If the camera is too busy, keep one movement — a zoom, a turn or a tracking shot — and try again.

Before you generate

1. Have you picked the right input: text, an image or supported references?
2. Does the prompt make the subject, action, camera and mood clear?
3. Does the aspect ratio suit the place you will publish the clip?
4. Have you checked the audio options and estimated credits?
5. Do you have permission to use the people, images, music and brand assets you upload?

For a feature overview, visit the Seedance 2.5 guide, or go straight to AI video generation.