Skip to main content
Requires Premium.
Text-to-Video generates a clip from a prompt alone. Use it when you want to move fast and don’t need precise control over the first frame composition.

​
How to use it

  1. Open the video generator and start or open a Session
  2. Select Text-to-Video
  3. Write a prompt describing the subject, setting, and motion
  4. Select a resolution
  5. Click Generate

​
Writing a prompt for Text-to-Video

A Text-to-Video prompt needs to cover both the scene and the motion in one description. Structure it as: subject → setting → action → camera. Good: a woman with long dark hair, standing in a sunlit bedroom, turning slowly toward camera, slow zoom in, soft natural light Bad: beautiful woman, high quality, cinematic — no motion, no setting, no camera direction. Without a first frame the model makes its own composition decisions. The more specific your prompt, the less the model improvises. See Video Prompting for a full breakdown of motion and camera terms.

​
When to use Text-to-Video vs Image-to-Video

Text-to-VideoImage-to-Video
SpeedFaster to startRequires a first frame
Composition controlLow — model decidesHigh — you set the frame
Character consistencyVariableConsistent with first frame
Best forQuick experiments, abstract scenesSpecific characters, precise shots
If character accuracy and composition matter, generate a first frame in Realism and use Image-to-Video instead. See Workflow.

​
Token cost

Text-to-Video uses tokens. See /premium for current rates.

​
Common issues

Text-to-Video has no reference frame, so character appearance can drift. If consistency matters, switch to Image-to-Video with a generated first frame.
The prompt doesn’t describe any movement. Add explicit motion terms — slow pan, walking forward, hair moving, camera pull back — to the prompt.
Without a first frame the model interprets composition freely. Add camera and framing terms to your prompt — close-up, mid-shot, wide angle, overhead — or switch to Image-to-Video for full control.

Image-to-Video

Animate a generated first frame for more control.

Video Prompting

Motion and camera terms that work.