Vixxxen / AI OFM platform / Seedance 2.5 guide

Model Guide

Getting the most out of Seedance 2.5

Seedance 2.5 makes the most lifelike clips in Vixxxen, and it is also the model that uses credits fastest. A few simple habits are the difference between two clips you love and two re-rolls. This is what we learned from nearly 500 Seedance 2.5 clips our customers made.

27 creditsfor a 5-second clip at 720p, the default

1 action, 1 camera moveper clip, plus the setting and the lighting

Image firsta photo is 1 credit, so nail the look before you animate

What a clip costs

Seedance 2.5 is priced by the second. A 5-second clip at 720p, the default, is 27 credits.

ClipCreditsWhen to use it
5 seconds, 480p Draft13.5Cheapest way to test a motion idea
5 seconds, 720p HD27The default, and the right first clip
10 seconds, 720p HD54Only once the 5-second version works
15 seconds, 720p HD81More than a whole Starter month
5 seconds, 1080p Full HD67.5For a finished, proven prompt

A Starter plan's 60 monthly credits cover about 2 clips of 5 seconds at 720p. If a clip fails or is blocked, its credits go back to your balance automatically.

Nail the look with a cheap image first

Seedance 2.5 always starts from a picture. The start image decides her face, the outfit, the setting and the light. The video only adds the motion.

A photo costs 1 credit with VIXN-2 (Seedream 4.5 costs about the same). One 5-second clip costs as much as 27 photos, so do your experimenting with photos.

  1. On the Image tab, pick your character and make 4 to 8 photos of the shot you want.
  2. Tap Enhance on the image prompt bar. Prompt Enhance turns a short idea into a rich, detailed photo prompt.
  3. Pick the sharpest photo, with her face clear, well lit and not tiny in the frame.
  4. Open its menu, choose Send to Video, and pick Seedance 2.5. That photo is now your start image.

Pick the right starting image for the video you want

Seedance 2.5 continues from your start image: the first frame of the clip is that picture. So the picture has to set up the move you want before you write a word.

In our data, clips that started from a picture made in Vixxxen were downloaded 45% of the time, against 33% for other start images, and none of them failed. Clips that started from a VIXN-2 picture: 57%.

The prompt recipe

One subject, one action, one camera move, the setting and the lighting.

The shape of a good prompt: Who moves + one action with a speed word + one camera move + the setting in a few words + the lighting.

Aim for 25 to 60 words. Longer is not better: in our data, prompts of 300 words or more were kept less than half as often as short ones.

Bad vs good Seedance 2.5 prompts

Three prompts that waste credits, and the versions that work.

1. Too vague

Wastes credits

“pose slowly and sensually”

No action and no camera, so the model guesses. You get a random sway, and the usual fix is a re-roll at full price.

Works

“She slowly turns her head toward the camera and gives a soft, knowing smile. Slow push-in from a medium shot to a close-up. Cozy bedroom, warm bedside-lamp glow, soft shadows.”

Why it works: One action with a speed word, one camera move, and the setting and light named in a few words. The model has nothing to guess.

2. Too much at once

Wastes credits

“She walks to the window, takes off her jacket, turns around, sits on the bed, lies back and blows a kiss while the camera circles her and zooms in and out.”

Five actions and three camera moves in 5 seconds. The model rushes, skips half of it, or bends her body to fit it all in.

Works

“She walks slowly toward the camera across the hotel suite, holding eye contact, her dress swaying with each step. Smooth tracking shot moving backward ahead of her. City lights through the window, warm lamp light.”

Why it works: One action that fills the whole clip and one camera move that suits it. Save the next action for the next clip.

3. Describing the photo instead of the motion

Wastes credits

“A stunning 25-year-old woman with long wavy chestnut hair, green eyes and full lips, wearing a red silk slip dress with thin straps and gold hoop earrings, standing on a balcony. Preserve her exact face, body and outfit. Ultra realistic, 8k, masterpiece...”

It says nothing about what moves, and every detail you repeat is a detail that can fight the picture. Faces drift when the words and the image disagree.

Works

“A light sea breeze lifts her hair and the hem of her dress as she leans on the balcony rail, gazing out at the water. Slow orbit from her side to the front. Sunny seaside balcony, golden late-afternoon light.”

Why it works: The picture already carries her face and the dress. The words only add what moves, so the clip stays true to the image.

Start short, and use Prompt Enhance

Make your first clip 5 seconds at 720p, the default (27 credits). Want to test a motion idea for less? 480p Draft is half the price: 13.5 credits for 5 seconds.

When a 5-second version looks right, keep the same image and prompt and only then make it longer or sharper.

Prompt Enhance lives on the image prompt bar, so use it where it counts: on the start image. Keep the video prompt short and about motion.

The common money-wasters

Motion that looks real

Keep her face the same

Fixing a bad clip

Find the symptom, change one thing, and test the fix at 480p Draft before you spend on 720p.

What went wrongWhat to change
Her face changedUse a closer, sharper start image, slow the motion down, and drop any head spins.
Too much is happeningCut the prompt down to one action.
The camera wandersName one camera move, or write "fixed camera".
A cut or a second scene appearsAdd "one continuous shot".
Unwanted music, captions or textAdd "no subtitles, no on-screen text, no background music".
Arms or body bend oddlyAsk for a slower, smaller movement.
The clip is right but too shortKeep the image and the prompt, and only change the length.

Using references: face photos and a motion video

What Seedance 2.5 accepts on Vixxxen

What references cost

Without a motion video, Seedance 2.5 Motion Control costs the same as Seedance 2.5: 27 credits for 5 seconds at 720p. Reference images add nothing.

With a motion video you pay for the video's seconds plus the clip's seconds, at 3.3 credits a second at 720p. A 5-second video for a 5-second clip is 33 credits; a 15-second video for the same clip is 66. Trim the video to the part you need.

When to use each

Good face references

Good motion videos

How to write the prompt

In the prompt, say what each reference is for, then describe the scene in a few words. The video carries the movement, so you do not need to describe it step by step.

“@Image1 is her face and @Image2 is her outfit. She follows the movement in @Video1, with the same framing as @Video1. Cozy bedroom, warm lamp light.”

An honest word on blocks

Be ready for blocks. In our data, Seedance 2.5 Motion Control jobs with a motion video were blocked 37% of the time (the credits came back automatically). The model's own identity filter is strict about realistic faces combined with a video of another person, and Vixxxen cannot switch it off.

When they went through, those clips were downloaded 55% of the time, one of the best rates we measured, against 19% for Motion Control without a video.

To get through: give it one job (she follows the movement in @Video1) rather than a face swap plus a new background plus lip sync in one go. If a job is blocked, do not re-run the same thing: change the prompt or the video.

Make your credits go further

A month that works: 8 photos to nail the look (8 credits), one 480p Draft to test the motion (13.5), then one 720p final (27). That is about 49 credits, with room left for a second try.

Making more than a couple of clips a month? Spark is $34 a month for 300 credits, about 11 clips of 5 seconds at 720p. Compare every plan in the app.

Frequently asked questions

How many credits does Seedance 2.5 use?

Seedance 2.5 is priced per second: 2.7 credits a second at 480p, 5.4 at 720p and 13.5 at 1080p. A 5-second 720p clip is 27 credits.

How long should a Seedance 2.5 prompt be?

About 25 to 60 words: one subject, one action, one camera move, the setting and the lighting. When you start from an image, describe the motion, not the picture.

Do I need a start image for Seedance 2.5?

Yes. Seedance 2.5 on Vixxxen is image-to-video: it animates a start image you choose. Make that image first with VIXN-2 or Seedream 4.5, which cost about 1 credit a photo.

Why did Seedance 2.5 change her face?

Usually the start image is too small or soft, the motion is too fast, or the prompt re-describes her and fights the picture. Use a closer, sharper image, slower motion, and a motion-only prompt.

Can I use reference images or a motion video with Seedance 2.5?

Plain Seedance 2.5 uses one start image and an optional last frame. Seedance 2.5 Motion Control takes up to 10 reference images (@Image1 to @Image10) and one motion video (@Video1, 2 to 30 seconds). With a video you pay for its seconds plus the clip's: 33 credits for a 5-second video and a 5-second clip at 720p.

What camera moves work best in Seedance 2.5?

One move per clip: a slow push-in, pull back, orbit, tracking shot, handheld, or a fixed camera. Stacking several moves in one clip makes it unstable.

How long should a Seedance 2.5 clip be?

Start with 5 seconds. Clips can run from 4 to 30 seconds, but a longer clip costs more and has more room to go wrong, so go long only once the 5-second version works.

Make your first Seedance 2.5 clip

Create a consistent character, nail the look with photos, then bring the best one to life. Every paid plan includes every image and video model.