Images and video

When your library has no picture that fits, you can have one made. Archetype generates images with GPT Image 2.5 and video clips with Seedance 2.5, from a sentence and, if you like, from your own photos. Everything is paid for in AI credits, and the price shows before you send.

There are two places to do it, and they differ in what is added to your prompt and where the result goes:

To use either, you need permission to edit the Library. Without it, or if generation isn't switched on for your workspace, these controls don't appear.

Make an image or a clip on Create#

1. 1

   Switch the mode

   On Create, open the first chip in the prompt box and pick Image or Video instead of Design.

2. 2

   Describe it

   Write what you want to see. Nothing is added to it: no brand rules, no direction. Type @ or press the photo button to work from photos in your library. An image can use up to 4 of them as references (a product to place, a scene to match). A video opens on the first photo you pick.

3. 3

   Choose the settings

   Set the chips along the bottom of the box (see the table below). Each option shows what it costs.

4. 4

   Generate

   Press the arrow. The picture or clip appears on the wall while it is being made. A video takes a few minutes.

The wall shows designs, images and videos together, newest first. Use All, Designs and Images & video at the top to filter.

Keep it#

What you make on Create stays on Create until you keep it. Hover a card and press Save to assets to copy it into your library, where agents and the rest of the team can use it. A saved card says In assets. You can also open it and press Download.

Deleting something from Create doesn't touch the copy in Assets.

Ask for a new version#

Open an image or a clip, and use the Chat tab beside it to say what should change: "warmer light", "lose the second chair", "slower camera". Each change is a new version, and every version stays in the thread to go back to.

• An image is edited rather than redrawn, at the same size and shape.
• A clip can't be edited, so it is rendered again from what was asked for it, with your change added. It keeps the same opening photo and length.

Changes start from the version you have open. On the wall, a line of versions shows as one card.

Generate into the library#

On Assets, press Generate. The drawer has an Image and Video switch, a box for your description, and the same settings as Create. Add photos from your library to work from; as you type, the drawer suggests ones that seem to match.

Here, Claude writes the brief the model gets from your brand (its imagery rules, palette and the photos you picked), so the result looks like your brand's photography rather than stock. It never adds photos you didn't pick. The result is added to your library as soon as it is made, in the folder you were in.

On any image in your library, Generate from this opens the same drawer with that picture already picked.

Settings#

Price follows the pixels for images, so 4K costs about 8 times 1K, and the length for video. The exact number of credits shows on each option and on the button. If something fails, no credits are used.

Photos inside designs#

Two more places make a photo for you while you design:

• In a template, a Create design or a component example, every photo field has Or describe a new photo…. Describe it and the picture is made and placed in that slot.
• On Create, Generate photos when none fit (under Photos in the Brand chip) lets Claude have a photo made when nothing in your library suits the design.

Both add the new photo to your library.

A campaign can hold images and videos as pieces too, made in the campaign's look and kept on Create like the rest until you save them.

From an agent#

Connected agents can do the same with the generate_media MCP tool, when the library has nothing that fits. Images come back when they are made; for a video, the agent gets an id and checks back until the clip is ready. Results land in your library with a link the agent can use. See MCP tools.

Next steps#