Making a video from text alone
No picture needed — describe the scene and the model films it.
Models tagged Text→video make a clip from your prompt alone. Pick one in the Video lane and the Source image panel goes away.
Make a clip
- 1Click Video in the sidebar.
- 2Open Model and pick a model tagged Text→video. To list only those, type “text” in Search models….
- 3Pick a Length and, if the model offers one, a Resolution.
- 4In Prompt, describe the scene, the subject, the motion and the camera.
- 5Click Generate video. The clip lands in the result panel and your Library.
Write the whole scene
With no picture to start from, the prompt does everything: who or what, where, what happens beat by beat, and how the camera moves. “A red kite lifts off a beach at dusk, camera tilting up as it climbs, waves rolling in below.”
✨ Direct isn’t offered here. It writes motion from a picture, so it’s only on picture-to-video models.
Shape and sound
Text→video clips come out widescreen (16:9). There’s no shape picker. For a vertical clip, make a 9:16 picture in the Image lane, then animate it.
Veo 3.1 Fast, Veo 3.1 Lite, Kling 2.6 Pro and Grok Video 1.5 Lite make sound with the clip, and the price includes it. PixVerse V6 clips are silent.
Picking a model
Avoid and entities
PixVerse V6, Kling 2.6 Pro, Veo 3.1 Fast and Veo 3.1 Lite have an Avoid (negative prompt, optional) box. List what you don’t want, like “text, morphing faces”.
Type @ in the prompt to name a saved character, prop or place. Text→video models get its written description. To keep a face or product exactly the same, use Kling O3 · Entities or Veo 3.1 Fast · Entities instead. They build the clip from your entities’ pictures, so name at least one @entity in the prompt. See Entities.
Text first or picture first?
Text→video is fastest when you have no picture yet. When the look matters, two steps usually win. Make a picture until the frame is right — that’s cheap to repeat. Then animate it once.