Elements are reusable building blocks you can drop into a prompt — a specific character, product, style, or background — so the AI uses that exact saved thing instead of guessing from words. You create an element once (by uploading images or by turning an existing model/content into one), then reference it again and again across the tools that support prompts.

An element is a saved, reusable item you reference by name in a prompt. Each one has a frontal image (its main reference), a name (required), an optional description, and a category that tells the app what kind of thing it is — Character, Product, Style, Background, or Others. Once saved, an element shows up as a colored chip inside your prompt and appears in the element picker for next time.

You work with elements from inside a tool that takes a prompt (for example image-to-video, edit-video flows, and storyboard shots). Open the element picker — using the Add element (or Add elements) button — to browse what you already have, reuse community elements, or create a brand-new one. For how the chips behave inside the prompt itself, see Using elements in a prompt.

From the prompt, tap Add element (or Add elements) to open the picker. It opens full-screen with two tabs at the top — Personal and Community — and a row of filters underneath. If you've never made an element, the picker opens straight to the Community tab so you have something to browse right away; otherwise it starts on Personal.

Selecting an element (personal or community) inserts that element directly into your prompt. Picking an Influencer (model) or Content item creates a personal element copy from it and inserts that.

Below the tabs, three filters narrow what the grid shows:

This filter row is hidden in flows that are locked to a single kind of thing (for example the storyboard "Add character" flow, or the product flow), since there's only one category to show.

In the picker's Personal tab, tap the Create Element card (the dashed card with a plus and an Add a new element button). That opens the create form. Fill in at least a frontal image and a name, then tap Save. Your new element is created and immediately selected into your prompt.

No. The picker is a browseable grid that loads more items as you scroll, but there's no text search — you find elements by scrolling the Personal or Community grid (or by switching filters).

In the create form, drag and drop an image onto the upload area or click it to choose a file — the prompt reads Drag and drop an image here, or click to upload. The first image you add is the frontal image; the Add additional images (optional) area appears once you've added the frontal one. Each image uploads with a live progress percentage, and you can remove any image before saving.

You add a frontal image plus up to 2 additional images of the same element. Once you've added two, the option to add more disappears. (An on-screen hint may mention "up to 3", but the form only accepts two additional images — so don't prepare a third expecting to upload it.)

Keep the name to letters, numbers, spaces, and the characters - _ ( ). Other special characters aren't allowed — you'll see "Special characters aren't supported in the element name yet." and the Save button stays disabled. Names must also be unique within your set; if you reuse a name you'll get "An element with this name already exists. Please use a different name."

Save stays disabled until you've added a valid frontal image and typed a name that isn't empty and contains no unsupported characters. Once both are satisfied, Save lights up. The description and additional images are optional and never block saving.

When Public is on, your element is visible in the Community tab and other people can reuse it. Turn it off to keep the element private to your account. Making something private is a paid feature: if you're on the free plan and try to switch a public element to private, you'll be prompted to upgrade instead.

Five categories describe what kind of thing the element is:

The category appears as a small tag in the prompt picker. On the element's detail card, the category tag is shown only for other creators' elements — on your own elements the @name appears but the category tag is hidden. In flows that are locked to one category (like the storyboard Elements column), the category is fixed and the dropdown is hidden.

Yes. Switch the picker filter to Influencer to pick one of your portrait Models, or Content to pick a finished image. Tapping it creates a personal element from that model or content (named after it) and selects it into your prompt. These are created as private to you.

The element form is built around images by default. Video and audio element types (with their own upload areas and a small Image / Video / Audio switcher at the top of the create form) only appear when that capability is enabled for your account; otherwise you'll only see the image flow.

Open the element's detail card and tap the small edit (pencil) button next to its name, then change the name, category, public status, or description and hit Save. Editing is only offered for elements you uploaded yourself — elements created from a model or from content can't be edited this way. When editing, the images themselves are locked; you change the text details and visibility, not the uploaded files.

Tapping the small info button on an element (or opening one from a prompt chip) brings up a detail card showing:

For video or audio elements, the card plays the clip with standard controls instead of showing the image grid.

To delete a saved element from your library for good, go to History and open its element card (its ⋯ / more actions menu) and choose Delete element. You'll be asked to confirm — "Delete element" / "Are you sure you want to delete this element?" — and once you confirm, the element is removed from your account. This is different from removing an element from a prompt (below), which only clears it from that one prompt.

Inside a prompt, you remove an element by deleting its chip in the prompt area. The app warns you that "Deleting this element also clears its references from the prompt," so any place you referenced that element in that prompt is cleared too. This doesn't delete the saved element from your library — it only takes it out of the current prompt.

Once an element is saved, you reference it by inserting it from the picker — it appears as a colored chip showing its @name and thumbnail, sitting inline with your text. That tells the model to use that exact element at that spot. You can mix several elements with normal prose in a single prompt. See Using elements in a prompt for chip behavior, limits, and how mentions combine with your model and voice.

It depends on the tool and quality tier you're using — there isn't a single fixed limit. For example, Generate Image's Chat to generate unlocks more element slots at higher quality (4 at Ultra, 8 at Ultra 4K, 10 at Ultra S), while other modes offer fewer. When you reach a tool's limit, the picker won't let you add more until you remove one. See Using elements in a prompt for the per-mode behavior.

Keep an element's look in step with the model you're generating with. Mixing mismatched styles — for example pairing a realistic portrait model with a cartoon-style element — can give inconsistent or warped results. For the most coherent output, pick elements whose style fits the model you've selected.

If you tap Save while an additional image is still uploading, you're asked "Some images are still uploading. Save the element anyway?" — choose Continue to save without waiting (the unfinished image is dropped) or Cancel to wait. If an image failed to upload, you'll get a similar "Error uploading images" prompt offering the same choice.

The picker and create form open as full-screen takeovers rather than centered dialogs. The element picker for prompts appears as a slide-in drawer titled Select influencer / element, grouped into an Influencer section (your portrait Model and any Voice model) and an Element section listing your elements with their category tags — tap any row to insert it. Creating and editing elements use the same fields as desktop, just stacked for a narrow screen.

Related: Using elements in a prompt · Creating a portrait model · Composer and prompt · Credits, plans & billing