> ## Documentation Index
> Fetch the complete documentation index at: https://www.krea.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Elements: reusable characters, locations, and objects

> Save a character, location, or object as an element with reference images and guidelines, then mention it with @ in Seedance Studio for consistency.

An element is a subject you want to keep consistent across generations: a character, a location, or an object. It bundles up to eight reference images with written guidelines under one `@tag`. Attach the element once and every image and the guidelines travel with your prompt.

<Note>
  Elements are available in Seedance Studio today, in **Video** mode. They are coming soon to all other Krea tools, so anything you save now will be ready to use there.
</Note>

<a href="https://www.krea.ai/seedance-studio" className="not-prose group my-6 flex items-center justify-between gap-4 rounded-xl border border-gray-200 dark:border-gray-800 px-5 py-4 no-underline transition-colors hover:border-gray-300 dark:hover:border-gray-700">
  <div>
    <div className="font-semibold text-gray-900 dark:text-gray-100">Open Seedance Studio</div>
    <div className="mt-0.5 text-sm text-gray-500 dark:text-gray-400">Elements live in the References panel in Video mode.</div>
  </div>

  <span aria-hidden="true" className="text-gray-400 transition-transform group-hover:translate-x-0.5 dark:text-gray-500">→</span>
</a>

## Create an element

<Steps>
  <Step title="Open References">
    In Seedance Studio, switch to **Video** and click **References** in the composer. The panel lists everything you have saved under **All**, **Characters**, **Locations**, and **Objects**.
  </Step>

  <Step title="Start a new element">
    Click the **New element** tile. Starting from a category tab sets the category for you; starting from **All** lets you pick it before you save.
  </Step>

  <Step title="Name and tag it">
    Give the element a name, such as `Maria`. The `@tag` you will type in prompts is derived from the name, so `Bolex camera` becomes `@bolex-camera`. Tags use letters and hyphens only and must be unique among your elements. You can edit the tag by hand.
  </Step>

  <Step title="Add reference images">
    Upload up to eight images of the same subject. Shots from different angles, distances, and framings give the model the most to work with. The first image is the cover shown in the panel; right-click any image and choose **Make this the cover** to change it, or use **Remove image** to drop one.
  </Step>

  <Step title="Write guidelines">
    Describe what the subject is and what must stay the same in every shot, up to 400 characters. For example: `Woman in her early thirties with a sharp jet-black bob, pale skin, and deep burgundy lipstick. Keep the haircut and the black turtleneck in every shot; lighting and setting can change.`
  </Step>

  <Step title="Save">
    Click **Create character**, **Create location**, or **Create object**. The element appears in its category and in **All**.
  </Step>
</Steps>

## Use an element in a prompt

Click an element tile in the References panel, or type `@` in the prompt and pick it from the **Elements** group. Either way, Studio attaches all of the element's images as references and places one chip in your prompt, for example `Maria walks slowly into the empty warehouse`.

When you generate, the model receives the element's images, and the prompt names the subject and includes your guidelines so the model knows which images belong together:

```text theme={null}
Maria walks slowly into the empty warehouse. Maria: the character shown in @Image1, @Image2 and @Image3. Woman in her early thirties with a sharp jet-black bob ...
```

To detach an element, delete its chip from the prompt, click its tile again, or use **Reset all** in the panel. The count badge on the **References** button shows how many references are attached; hover it for a breakdown by element.

## Reference limits by model

Each video model accepts a limited number of reference images, and elements count toward that limit. Flux 3 takes only a start and end frame, so elements cannot be attached while it is selected.

| Model                             | Image references | Total references |
| --------------------------------- | ---------------- | ---------------- |
| Seedance 2.5                      | 30               | 50               |
| Seedance 2, 2 Fast, 2 Mini        | 9                | 12               |
| Hailuo 3                          | 9                | 15               |
| Gemini Omni Flash, Omni Flash 1.1 | 10               | 11               |
| Flux 3                            | None             | None             |

If an element has more images than the selected model has room for, Studio asks before attaching it: **Remove other references, or truncate this element to X images (down from Y)**. Confirming attaches the first X images in the element's order, which is another reason to put your best shots first.

<Tip>
  Switching to a model with fewer reference slots drops references it cannot take, including an element's images. Reattach the element after switching if you still need it.
</Tip>

## Edit or delete an element

Hover an element tile and click the pencil to edit it. You can rename it, change its tag or category, rewrite the guidelines, add images, remove images, or pick a new cover. Saving an element that is currently attached updates the attached references as well.

To delete an element, open it and click **Delete**. Its images stay in your assets; only the element is removed.

## Tips for consistent results

* Use shots of the same subject from several angles and distances rather than eight near-identical frames.
* Keep guidelines about identity: features, wardrobe, materials, or architecture that must not change. Leave action, lighting, and setting to the prompt.
* Mention the element where it acts in the sentence. Seedance orders references by where they are first mentioned.
* Give each element a short, distinct tag so `@` suggestions stay easy to scan.