Text to video
Every control on the text-to-video generator, and what it costs you.
Text to video is the default mode: you supply a prompt, the model supplies everything else. This page covers each control. For the prompt itself, see Writing prompts.
Prompt
Up to 5,000 characters. The model reads the whole thing, so use the room — a prompt that names the subject, the action, the camera, and the light will beat a one-line description nearly every time.
Model
| Model | Max length | Resolutions | Good for |
|---|---|---|---|
| Seedance 2.5 | 30s | 480p · 720p · 1080p | Final shots, long takes, real people |
| Seedance 2.5 Lite | 15s | 720p | Faster turnaround at one quality |
| Seedance 2.0 Pro | 15s | 480p · 720p · 1080p · 4K | When you need a 4K master |
| Seedance 2.0 Fast | 15s | 480p · 720p | Drafting |
| Seedance 2.0 Mini | 15s | 480p · 720p | Cheapest drafting |
Switching models changes which options below are available — the generator only shows what the selected model actually accepts, so the panel will shift as you change model.
The Seedance 2.x family requires a Standard or Premium plan. On a Starter plan or without a subscription, selecting one opens the upgrade dialog instead.
Length
4 to 30 seconds on Seedance 2.5, 4 to 15 on the rest. Credits scale with duration, so a 20-second clip costs roughly four times a 5-second one at the same resolution.
Long clips drift. If a 20-second prompt keeps losing the subject halfway through, you will usually get a better result generating two shorter shots and cutting them together.
Resolution
480p, 720p (default), or 1080p; Seedance 2.0 Pro adds 4K. This is the single biggest lever on cost — 1080p costs several times what 480p does for the same duration.
The workflow that wastes the least: iterate on the prompt at 480p and 4 seconds until the shot is right, then re-run the finished prompt at your target resolution and length.
High resolutions are also plan-gated. Without a Standard or Premium plan you get the lowest tier the model offers.
Aspect ratio
16:9 (default), 21:9, 4:3, 1:1, 3:4, 9:16, or adaptive. Pick the
one your destination wants — 9:16 for vertical social, 21:9 for a widescreen
look. adaptive lets the source decide, which matters mainly in
image to video.
Audio
On by default. The model generates a soundtrack — ambience, effects, and speech where the prompt implies it — together with the picture. Turn it off if you intend to score the clip yourself.
References
Beyond the prompt, you can attach:
- Reference images — up to 30, 10 MB each. Use these to carry a character, a costume, or a colour palette into the shot.
- Reference video — up to 10 clips, 200 MB and about 30 seconds each. Use these for motion and camera behaviour.
- Reference audio — up to 10 files, 15 MB and about 30 seconds each, to drive timing or a voice.
Attaching a source video changes how the generation is priced: billing includes the input footage, with a 4-second minimum. The cost shown before you submit already accounts for this.
Exact limits come from the selected model, so a different model may allow fewer references than the numbers above.
Cost
The credit cost updates live as you change model, resolution, length, and inputs. The number displayed before you submit is the number that gets charged — the same formula runs on the server. To compare plans against your expected settings, use the calculator on the pricing page.
After generating
The task runs on the server, so you can close the tab. Finished clips land in the results panel and in My Assets. Each run is an independent sample: regenerating an unchanged prompt gives a different take, which is often the cheapest fix for a shot that was nearly right.