Ask the Agent for a clip and have it placed for you

The Video tab gives you every control. The Agent gives you a shortcut: say what the clip shows and where it belongs, and it handles the rest. This guide covers how to word the request and what to expect while the render runs.

Say what and where

Open the Agent from the top bar and describe one shot together with its position. A few examples:

• "Generate a vertical clip of steam rising from a coffee cup and put it first."
• "Make a clip of rain on a window and add it after clip 3 when it's done."
• "Animate @i2 so the clouds drift slowly."

A plain "make a video of" request produces one clip. A planned, narrated production with several shots is a separate kind of request that you name as a story or an episode.

The Agent sets the shape of the clip (vertical or widescreen) from your project. The model decides the clip length. For control over the model, resolution, duration and number of takes, use the Video tab.

What you see while it renders

1. The Agent starts the job and replies right away.
2. A card appears in the chat showing your prompt and Rendering… with a running clock. A line near the bottom reads Rendering a video….
3. The render runs in the background. Keep editing, or ask the Agent for something else.
4. When it finishes, the card turns into the clip. Hover to play it, use Expand for a larger look, and open … (More actions) for Add to timeline, Add to Media and Add to Library.

Placement follows your words. The clip reaches the timeline only when you ask. If you did, it drops in by itself the moment it finishes, at the spot you named, such as first. If you did not, it waits on its card until you drag it in or choose Add to timeline.

The render spends credits. A Stop button replaces Send while the Agent is working, and a failed render shows its error on the card with a Retry link.

Let it wait for the render

When the next step depends on the clip, say so in the same message: "Generate the clip, and when it lands, add a title over it that says Opening Night." The Agent can wait for running renders to settle and then finish the job in the same turn. If a render is still going after a while, it tells you how long it has run and waits for your go-ahead before checking again.

Start from a picture

Mention an image or a clip with @ and the Agent anchors the render on it.

• By default it first designs an opening frame from your reference with the image model. That frame appears as its own card, and the video then animates that exact frame.
• Say "animate this image" and your picture becomes the opening frame unchanged. The prompt then needs to describe only the motion.
• An audio reference contributes what it carries, such as spoken words or the mood of the sound, to the prompt.

If the model refuses a picture, the render steps down on its own, first using the picture as a looser reference and then, unless you asked to animate that image, using the text alone.

Review before you build on it

Play the finished clip from its card and check the first and last second, the faces and the audio. Then ask for the next step, such as "trim it to the strongest moment" or "add a fade out". If it is not right, ask for a new version with one change to the prompt.

Ask for one now

Open DepCut, press Cmd+J or Ctrl+J and ask the Agent for your next shot and its position.