Generate video

Ask Agent to generate video. Discuss storyboards, shot order or a larger narrative first when needed, or use a video Skill.

Prepare inputs and shot requirements

Start from text or an approved design image. Resolve structural and perspective errors in images first because they may carry into the video.

Describe subject movement, camera motion and stable elements: “Slowly approach the entrance; preserve the building profile and show people walking naturally.” Focus a short clip on one action for easier review.

Generate

1. Request video and preferably retain default models and parameters.
2. Attach the needed images, video or other inputs and explain their roles. Follow input-limit prompts.
3. Specify ratio, resolution, duration and audio when delivery requires them.
4. Wait for processing, send and monitor the original conversation.

Upload limits and model input limits both apply. Adjust duration, dimensions, counts or combinations according to the task's prompt.

Review the full clip

Play it from beginning to end. Check drifting building profiles, abrupt camera jumps, unusual people/objects, duration and aspect ratio. Listen to the full audio if present.

For changes, reference inputs and output and identify the segment and issue. Regenerating a clip is different from editing an existing video; changing a prompt does not guarantee every other frame remains identical.

Download the video asset when satisfied. A static cover in a PDF is not a playable video deliverable; provide the video file separately.

Video director · Quick video browsing · Download results · Documentation