What is a multishot AI video, and when to use one?

Updated 2026-10-27·3 min read·Length & structure

One take, several cuts

A normal AI video clip is one shot: one camera, one framing, a few seconds. A multishot is one take with cuts inside it. You write a scene as a list of beats, and the model renders them as a single clip that cuts from setup to setup. In ShowMaker Studio a multishot can run up to thirty seconds on Wan 3.

Our episode, Marry My Broken Son, is about 67 seconds long and built from eight multishot panels. Each panel is a scene's worth of coverage in one render.

How to write one

In Scene Flow, the bar between panels adds an action, a dialogue line, a single shot or a multishot. Then:

1. Make each beat one camera setup. Say what happens, and tag the people, rooms and props by their @tags.
2. Put the line in quotes after the speaker. The quoted line becomes the voice line for that beat.
3. Give the first beat a start frame. Write what the opening instant looks like, then press Draw. It paints the frame from the cast and room sheets.
4. Render. The take plays beside the panel, with the references it used underneath. Add beats to grow the take, then render again.

When a multishot is the right call

Use it when a scene is a run of connected moments in one place, with the same people. The cuts come from one generation, so the room and the light tend to hold across them.

Don't use it when a beat is quietly holding two setups. The panel warns you when one beat is trying to be two camera setups. A beat that says "cut to" is the usual culprit. Press Split and each setup gets its own beat.

Watch the other warning too: speech that won't fit. We had a beat with six seconds of speech in a five-second slot. The line gets cut off or rushed. Lengthen the beat or shorten the line before you render.

When one part of the take is wrong

This is the main downside of a multishot. Everything renders together, so a mistake in one beat sits inside a take you otherwise like.

In our episode, Tristan is in a wheelchair, and his rise belongs to the next scene. The model stood him up at the end of the take anyway. We had two options that didn't involve rendering the whole thing again:

• Trim it. We dragged the trim handle back so the take ends before he stands. The trim is kept and nothing is spent. That left the take shorter than the lines it was meant to cover, so we added those lines underneath as single shots.
• Use Split into two shots. This breaks the multishot apart, so you can fix one shot without rendering the rest again.

Then Render scene cuts the trimmed take and the new shots together into one scene.

The short version

A multishot is cheap coverage when the beats are clean. Keep one camera per beat and keep each line inside its beat's length. Read the warnings. When something goes wrong, trim or split instead of rendering the whole take again.

For more on scene length, see long AI video scenes and Scene Flow.

The full walkthrough is in the guide, and the wheelchair fix is the Short The AI made him stand up too early.