Model
Select Seedance 2.5, not Seedance 2.0 or Mini.
How to Use Seedance 2.5: Create Your First AI Video
Start with Text to Video. The steps below cover Image to Video and Multi Reference after that.
If you are not signed in, Topview asks you to sign in before the generator opens.
Follow one path
Switch between Text to Video, Image to Video, and Multi Reference. For your first video, stay on Text to Video.
Set these once
Duration
Choose 30s for a complete first clip.
Resolution
Start at 720p, the trial resolution. 1080p is the highest setting shown in the generator. 4K is not a native export.
| Control | What to do |
|---|---|
| Model | Select Seedance 2.5, not Seedance 2.0 or Mini. |
| Duration | Choose 30s for a complete first clip. |
| Resolution | Start at 720p, the trial resolution. 1080p is the highest setting shown in the generator. 4K is not a native export. |
30s at 720p is the trial tier. Trial credits and the exact settings on your account are shown before you generate.
Text to Video
Use this when you do not have a picture yet. This is the full click path for a first Seedance 2.5 clip.
- 1
Open Seedance 2.5
Open Topview and go to the page, or open AI Video Generator and pick Seedance 2.5 from the model list. The Seedance 2.5 page has the model overview and the full parameter list. This guide stays on the click path. Check the model label before you do anything else. It must say Seedance 2.5, not Seedance 2.0 or Mini. If you are not signed in, Topview asks you to sign in before the generator opens.
- 2
Select Text to Video
Pick Text to Video. It is the fastest way to a complete clip when you do not have a picture yet. Leave the reference area empty. You do not need to upload an image for this first video. Add a file later only when a product or person must match a picture you already have.
- 3
Set 30s, 16:9, and 720p
Set Duration to 30s, Aspect ratio to 16:9, and Resolution to 720p. 30s and 720p are the trial tier. 1080p is the highest setting shown in the generator. 4K is not a native export. Your account shows the trial credits and the exact settings before you generate. 16:9 applies to Text to Video, so you can review the clip on a desktop.
- 4
Paste one example prompt
Go to the example prompts below and paste one into the prompt box. Start with the perfume ad, or try the action scene if you want a different type. Keep the three time ranges. On this first run, do not rewrite the whole prompt. Your goal is one complete 30-second clip. If you write your own prompt later, keep this shape: subject, action, setting, camera, lighting, and three beats.
- 5
Image to Video
Use this when the product or person must match a picture you already have. Keep the model on Seedance 2.5, duration at 30s, and start at 720p.
- 1
Select Image to Video
In the Seedance 2.5 generator, switch the mode from Text to Video to Image to Video. Check the model label again. It still needs to say Seedance 2.5, not 2.0 or Mini.
- 2
Upload one image
Upload one image. The frame follows that image, so do not force 16:9 when the picture is a different shape. One image is enough for this run.
- 3
Write the motion and the camera
In the prompt box, describe what moves and how the camera moves. Do not describe a different subject from the one in the picture.
- 4
Generate and change one sentence
Click Generate and watch the preview. If the subject does not match the image, change the motion sentence only, then generate again. Stay at 720p until that clip is worth keeping.
Multi Reference
Use this after one image already works and a second file has a different job. Keep the model on Seedance 2.5.
- 1
Select Multi Reference
Switch the mode to Multi Reference. Confirm the model label still says Seedance 2.5.
- 2
Give each file one job
Upload one file per job. Add a second file only when it adds something the first file does not, such as a background or a piece of audio.
- 3
Name each job in the prompt
Write the job in the prompt box. For example: “Image 1 is the product only. Image 2 is the background only.”
- 4
Generate, or go back to one image
Click Generate. If the files conflict, remove the extra file and return to Image to Video. Change one file or one sentence at a time.
Example prompts and results
These two Text to Video prompts are ready to paste. Each card is followed by the clip that belongs to that prompt. Neither needs to look like a finished ad.
Luxury perfume ad
perfume.txt
Create a stunning 30-second cinematic luxury perfume ad featuring an elegant amber glass bottle with a black cap. 0–8s: Golden liquid droplets fall in slow motion, creating beautiful splashes against a dark background. 8–18s: Reveal the perfume bottle on a glossy black surface. Slowly orbit around it, capturing glowing golden reflections, realistic glass textures, and dramatic lighting. 18–30s: Transition to a breathtaking hero shot with soft golden backlighting, subtle mist, and a slow camera push-in. Leave empty space for a logo. Ultra-realistic, premium commercial quality, elegant black-and-gold aesthetic, smooth cinematic transitions. Keep the bottle consistent. No subtitles or text. Soft ambient music only.
Continuous action scene
action.txt
Create a 30-second photoreal cinematic action scene of soldiers moving through a dusty, sunlit stone city street. 0–8s: A soldier in tan desert camouflage, helmet, and tactical vest presses against a stone wall and aims a rifle down the street. He fires as warm backlight and dust fill the air. The camera pushes in, then whips wider to a fireburst, a wrecked dark car, and other soldiers taking cover. 8–18s: The squad advances along the building. One soldier covers a dark doorway while another moves past a damaged sedan and sandbags. Handheld tracking. Keep the uniforms, helmets, and street consistent. 18–30s: Move closer as a soldier crouches beside the wall, then open to a wide shot: a vehicle explodes in a rolling orange fireball at the end of the street, with dust and fallen figures in the foreground. Photoreal, tense, 16:9. No subtitles or on-screen text. Low cinematic score only.
Fixes for a weak first clip
If the first Seedance 2.5 render misses, check these before you change the whole prompt.
Trap 1 — Your prompt is too vague
Bad: “Make a cinematic video.” What you get: empty motion, no story.
Fix: use the three time ranges in one of the example prompts.
Trap 2 — One action for 30 seconds
The model fills the dead time with waiting or repetition. A 30-second clip feels empty when one action has to last the whole duration.
Fix: give it three beats, such as 0–8s, 8–18s, and 18–30s.
Trap 3 — Too many references on run 1
Identity conflicts make it hard to tell whether the prompt even worked.
Fix: leave references empty on Text to Video. Add one image only after you need the subject to match a picture. Add a second file only when it has a different job.
Trap 4 — Rewriting everything after one bad render
You throw away the lines that already worked.
Fix: change one sentence, one timestamp, or one setting, then generate again.
Trap 5 — 1080p comes before the story works
A higher resolution does not fix an empty story.
Fix: finish the first clip at 720p. Move to 1080p only after that clip is worth keeping. 4K is not a native export.
When to stop tweaking and ship
Export your first Seedance 2.5 clip when all three are true:
- The 30s story reads clearly from start to finish
- The subject stays recognizable
- The ending has a few clean seconds for a logo
Add another file only when the first result already matches. Longer prompts are in the Seedance 2.5 prompt library. A different look can start from an example.
Generate your first Seedance 2.5 clip
Open Seedance 2.5 in Text to Video and paste one example prompt.