Describe a scene, not tags
Write a sentence ("a lone hiker on a ridge at golden hour") instead of comma-separated keywords.
Text → Video
Text to Video from a prompt
No photo required. Describe the scene, the camera and the mood, and the Seedance 2.5 model renders it straight into a cinematic, watermark-free clip.
What it does
Words in, motion out
Text to video skips the upload step entirely. You write a natural-language description of a scene and the model builds the footage from scratch — subject, environment, camera move and lighting all generated together.
It is the fastest way to test an idea when you do not have source art yet, or to spin up B-roll, intros and concept shots that never needed a real photo to begin with.
Prompting
How to write a prompt that lands
Seedance 2.5 reads natural language better than keyword lists. A few habits go a long way.
Lead with the subject
Put the most important thing first, then the action, then the environment and mood.
Add camera language
Phrases like "slow dolly-in, 85mm, shallow depth of field, 24fps" steer the look hard.
Name the motion
Strong verbs — drifting, surging, spinning — tell the model what should actually move.
Examples
Clips generated from text alone
No source image — every one of these started as a prompt.
FAQ
Text to video questions
No — that is the whole point of text to video. If you do have a starting frame, use image to video instead for tighter control over the subject.
Aim for 30–80 words: a clear subject, an action, the setting and a camera note. Too short is vague; a wall of tags confuses the model.
Yes — 16:9 for landscape, 9:16 for shorts and reels, plus 1:1 and others. Choose it before you generate.
Generate 4 to 10 second clips at 720p HD. Cost is flat across length, so a longer clip is better value per second when the motion suits it.
Explore
Keep creating
Start creating with Seedance 2.5
Turn a single prompt or reference image into a polished 4K clip in one shot — no editing timeline, no watermark, no forced subscription.