Choose reference mode or a starting frame
SuperSeedance currently accepts one image per image or reference request. Use reference mode when the image should guide the character in a new scene. Use image mode when the clip should begin from that image’s composition.
| What you need | Mode | Prepare |
|---|---|---|
| A character reference for a new shot | Reference to video | One clear image; describe the new action and setting. |
| Motion from a specific composition | Image to video | One opening frame with room for the planned movement. |
Prepare one reference you can reuse
Choose a single, clearly visible character. Keep the face, hairstyle, outfit and important accessories readable; avoid a collage, text overlay or a pose that hides the features you need.
Keep a copy of this image for every shot. Upload a JPG, PNG or WebP up to 10 MB, with width and height between 300 and 6,000 pixels and an aspect ratio between 0.4 and 2.5. Use an original or permitted image.
Keep identity fixed; change the action
Start with one short shot and a clear movement. Let the image show the character’s appearance. State the details that should stay consistent, then describe the action, setting and camera.
This is a template to adapt with your own image. It is not a promise that two separate generations will match.
Use the uploaded image as the character reference. Keep the face, hairstyle, teal jacket and body proportions consistent. One continuous five-second shot: the character walks from the left side of a quiet station platform to a bench and stops beside it. Eye-level side-tracking camera, full body visible, soft overcast daylight. Keep the same illustrated style throughout.Plan each shot before generating the series
Write a short shot list. Keep a fixed note for the character, outfit and style; give each shot its own action and camera direction. Attach the original reference again for every new request.
If you reuse a previous task’s settings, check the image field before submitting. Reusing settings does not guarantee the reference is attached. Match the model, resolution and aspect ratio when you want comparable shots.
Start with actions that keep important features visible. A distant subject, a rapid turn or a long occlusion gives you less to judge and more opportunity for visible changes.
Compare the whole clip with the reference
Watch the opening, middle and end, then compare adjacent shots. Check the features below before spending credits on the rest of the sequence.
| Check | Look for | Next adjustment |
|---|---|---|
| Face and hair | Changing face shape, hairline or accessories | Use a clearer reference and remove conflicting appearance instructions. |
| Outfit | Different colors, sleeves or patterns | Keep the same outfit visible in the source and name the key details. |
| Body and motion | Changing proportions, hands or contact with the ground | Simplify the movement and keep the subject large enough to inspect. |
| Style and lighting | A new rendering style between shots | Reuse the same style and lighting description. |
Questions and answers
Can Seedance guarantee the same character in every video?+
No. Reusing the same image and compatible prompts helps guide the results, but separate generations can still change details. Check each clip before assembling a sequence.
Can I attach several character images here?+
SuperSeedance currently accepts one image per image or reference request. Plan around that input instead of referring to attachments that are not present.
Does reusing settings carry the image into the next shot?+
Check the upload field. Reusing a task can restore its prompt and settings while asking you to add the image again.
Does a reference image guarantee the same voice?+
No. An image can guide appearance, not establish a voice reference. This workflow does not lock a speaker’s voice across clips.