Prompt for Sora
Generate a video of [subject, described specifically], doing [one clear action], in [setting, described with concrete detail]. Camera: [movement, e.g. "static locked-off shot" or "slow pan left to right"]. Lasting about [X] seconds. Keep the action simple and singular -- avoid combining multiple unrelated actions.
Getting a coherent first result from Sora comes down to three basics: one clear subject, one clear action, and a named camera movement — leaving any of these vague is what most often produces a clip that feels random or aimless.
What this produces
A short video following the subject, action, and camera movement you specify — a clean baseline result rather than Sora’s own default pacing and framing choices.
Variations
Two ready-to-copy versions for common starting points.
Generate a video of [subject, described specifically], [doing one clear action], in [setting, described with concrete detail]. Camera: [movement, e.g. “static locked-off shot” or “slow pan”]. Lasting about [X] seconds. Keep the action to one continuous movement.
Generate a video of [object, described in detail] on [a surface], with [camera movement, e.g. “slow rotation” or “close-up static shot”], soft even lighting, lasting about [X] seconds. Keep the object as the sole focus, with no competing background action.
See also: Prompt for Sora 2 and Prompt for Veo3 Multi-Shot Sequence.
FAQ
How long can a Sora video be?
This depends on your current access tier and changes over time — check current product documentation for the actual limit rather than assuming a fixed number.
Why does the motion sometimes look unnatural?
AI video generation is newer and less consistent than image generation — simpler actions and camera movements tend to render more naturally than fast or complex motion.
Should I use this or the Sora 2 version?
This is the general baseline structure — if you have access to Sora 2 specifically, the dedicated Sora 2 prompt covers its more layered scene-direction approach.
