Text to video, with a clear creative direction.
A good AI video prompt describes something a camera could capture. Give your subject one action, place it in a setting, and decide how the viewer should see it.
Studio preview · Generation and checkout are not open yet. Seedance integration is planned.
Build a prompt in five parts.
You do not need a screenplay for a short clip. Start with this compact recipe, then remove details that compete with the main action.
A subject gives the scene a center. An action tells the model what changes over time. The setting establishes context, while camera and light guide the look. Describe these directly: “a locked camera at table height” is more specific than “make it cinematic.”
For a first attempt, keep one subject, one action, and one camera movement. A short shot has limited room for entrances, transformations, cuts, and multiple locations.
Turn a broad idea into a filmable moment.
A beautiful morning coffee video.
A white ceramic cup sits beside a window on an oak table. A thin ribbon of steam rises gently. The camera makes a slow, small push toward the cup. Soft morning light from the left; the table and background stay still. One continuous shot.
The revision defines a subject, visible motion, camera path, and lighting. It is a writing example, not a promise that every detail will be reproduced exactly.
Choose settings to fit the action.
5 seconds: one simple beat
A ripple reaches the shore, a leaf turns in the breeze, or a camera reveals one object. Describe a single action that can read clearly in a short moment.
10 seconds: let a moment develop
Allow a slower camera move or a subject to cross the frame. Extra time does not automatically make several unrelated scenes coherent; keep the shot focused.
Choose 9:16 for a vertical composition, 16:9 for a wider scene, or 1:1 for a square composition. Keep subject placement in mind when changing formats. Resolution affects the credit quote, so check the pricing examples before submitting.
Review one thing at a time.
When a result becomes available, first check the subject and action. Then inspect the camera movement, edges, and changes in appearance. Revise the part that failed instead of replacing every sentence. A simpler action may help more than adding another paragraph of adjectives.
Accepted jobs continue asynchronously and appear in My Library. Closing the page does not cancel the job. Starting a revised prompt is a new generation with its own credit estimate.
Text-to-video questions
Should I write dialogue or sound effects?
These examples focus on visible action. Audio generation is not a promised Aura feature. If a scene needs speech, music, or precise sound timing, plan that work separately.
Can I use a reference image instead?
Yes, the studio also has an image mode. It is a better starting point when you already have a composition or subject appearance you want to guide the shot. Read the image-to-video guide.
Do longer prompts always work better?
No. Additional detail helps only when it clarifies the same scene. Conflicting camera moves, changing locations, and several actions can make the intended result harder to communicate.