How to make AI videos for TikTok and Reels

Generate at 9:16 from the start

The model composes movement and light across the whole frame. If you render 16:9 and crop to vertical, you throw away the sides where the composition lived and soften what remains. Set the format to 9:16 before you render.

Put the motion in the first second

• Open on movement — a push in, a reveal, a whip pan
• Keep one subject centred in the safe area, away from captions and UI
• Avoid slow fades from black; they read as dead air in a feed
• End near where you began so the clip loops cleanly

Prompt patterns that work vertically

Tall frames favour subjects with height and foreground depth: a person walking towards camera, a bottle rotating on a plinth, steam rising from a cup, rain falling past a streetlight. Ask for one camera move only — "slow push in", "gentle orbit", "tilt up" — and let the subject do the rest.

Length and pacing

Five to eight seconds per clip. If your idea needs longer, render two or three clips and cut them together rather than asking for one long take, which drifts. Consistent subject, three angles, one soundtrack is a far stronger post than a single wobbly ten-second render.

Turn a photo into a Reel

A product shot, a portrait or a landscape photo can become a vertical clip directly. Upload the photo, let expidvid scan it into a motion prompt, and it becomes the opening frame of the render — the subject stays recognisable while the camera moves around it.

Batch a week of posts in one sitting

Queue several drafts in fast mode, review them together, then re-render the two or three keepers in HD. That is a week of vertical content from one session, and every render stays in the gallery with a title and description you can reuse as a caption.