Make your first AI video

You can already make an AI image — now bring the same idea to life

If you have tried AI images, this may sound familiar: a single still image is not always enough, because your idea unfolds over time rather than in one moment. You may also have seen a short AI video on social media and wondered how it was made. A video generator turns a text description into a few seconds of video, much as an image generator makes an image. But your request needs one thing an image does not: what happens in the scene as time passes.

Your first video in four steps

1. 1

   Describe the scene and what happens

   Decide not only what appears in the video, but also what changes over a few seconds: does an object move, does the light change, does the wind pick up? This is the biggest difference from an image request — video needs an event over time, not just a description of a subject.

2. 2

   Specify the camera movement

   Say whether the camera stays still or moves: circles the subject, slowly moves closer or follows it from the side. Camera movement shapes the result as much as the subject itself.

3. 3

   Describe the rhythm and choose the duration with the control

   Say in the text whether the movement happens steadily or at one moment — for example, “steady, slow movement”. Video duration is not part of the request text: in most video generators, including Runway, you select it separately using a dedicated duration control, rather than writing seconds into the request.

4. 4

   Watch the result and assess the movement

   Compare the video with what you imagined: does the movement look natural, and does the scene do exactly what you asked? If something feels wrong, change just one thing next time — the movement, camera or duration.

Try this

Each example shows a different thing video can do that an image cannot: camera movement, a short event or a repeating loop. All three exclude people and text, because these are among the most common stumbling blocks for AI videos, just as they are for images.

An imaginary lighthouse on a rocky shore on a stormy evening, the camera slowly circles the lighthouse from left to right, waves crash against the rocks and the lighthouse light flashes steadily, no people, no text in the video.

An imaginary small paper boat floats downstream along an autumn ditch, the water carries it forward until it hits a boulder and tips over, the camera stays still throughout, no people, no text in the video.

An imaginary paved courtyard on a rainy day, drops of water trickle from the roof into a wooden barrel in a steady, repeating rhythm throughout the video, the camera stays still, a peaceful atmosphere, no people, no text in the video.

An imaginary garden swing in an empty yard in the afternoon breeze, the swing moves slowly back and forth, the camera stays still throughout, no people, no text in the video.

Refine the same video

Keep the same scene, lighthouse and atmosphere. Slow the camera’s circling movement to about half its current speed.

A focused refinement shows how one change — the camera speed in this case — affects the whole video. If you also want a longer video, adjust the duration separately using the tool’s own duration control, for example to about eight seconds.

When a real person appears in a video

Once you know the basics

Next, you can combine video with AI-generated speech to give the scene an audio track, or build a short sequence of consecutive clips into one scene. Both require the same care: describe each clip or audio scene as precisely as your first one.