> For the complete documentation index, see [llms.txt](https://docs.dreamerland.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.dreamerland.ai/image/moss-image.md).

# MOSS image

MOSS-Image-Base is the base text-to-image model in the **MOSS-Image** family from OpenMOSS. It supports natural-language prompts, tag sequences, and JSON layouts for controlling text placement and object arrangement.

What sets it apart is **layout control**: besides a normal prompt, you can draw **bounding boxes** that tell the model exactly where each person, object and piece of text goes.

> MOSS is in the **Create → Image** tab. Pick the **MOSS** family to open it.

![The finished example from this guide: every subject sits in the box it was given.](https://content.gitbook.com/content/tbrc6W2n35oVxftMovDQ/blobs/Tj7JPvMPiUEm9qhRoz49/07-result.png)

## Models

| Model              | Description                                   |
| ------------------ | --------------------------------------------- |
| **MOSS Image**     | The official MOSS Image model by OpenMOSS.    |
| **Moody Moss Mix** | A community MOSS Image merge by catlover1937. |

## What makes MOSS different

* **Layout control.** Draw a box and MOSS puts that subject inside it. You decide the composition instead of hoping the prompt lands right.
* **Text in the image.** A Text box renders the exact words you type, in the place you drew it, such as a title, a sign or a logo.
* **Characters and objects.** Mark a box as a Character to tell MOSS it holds a person or creature, so people in a group scene stay separate.
* **Colour palettes.** Give any box up to 5 colours to steer the colours of what is inside it.
* **Native 2K.** Generate a 2K image directly instead of upscaling afterwards.

## The MOSS tab

![The MOSS tab](https://content.gitbook.com/content/tbrc6W2n35oVxftMovDQ/blobs/1RpoAtunJ7EJSfgjXjBk/01-moss-tab.png)

* **Task:** Text to Image, or Image to Image to start from your own picture.
* **Prompt:** describe the whole scene. With bounding boxes, this is the overall caption of the image.
* **Styles:** add styles as with other models.
* **Default negative prompt:** on by default. Turn it off to write your own negative prompt.
* **Aspect ratio:** the bounding box canvas follows this ratio.
* **Bounding box:** opens the layout editor (see below).
* **Native 2K:** outputs a 2K image and uses double the credits.

## Bounding Box

Bounding boxes work with both **Text to Image** and **Image to Image**.

### 1. Open the editor

Tap **Bounding box** below **Aspect ratio**. Once you have drawn boxes, the card shows how many there are and a small preview. Tap **✕** on the card to remove them all.

![The Bounding box card with a preview of the layout](https://content.gitbook.com/content/tbrc6W2n35oVxftMovDQ/blobs/B2kHGX4ikXR24EfDq1ug/02-bbox-option.png)

### 2. Draw your layout

The editor opens a canvas in your current aspect ratio.

![The bounding box editor](https://content.gitbook.com/content/tbrc6W2n35oVxftMovDQ/blobs/g7rAZVscA6DVqWBpo5Tg/03-bbox-editor.png)

* **Draw:** drag on an empty part of the canvas, or tap **+ Add**.
* **Select:** tap a box, or its chip below the canvas.
* **Move:** drag the selected box.
* **Resize:** drag one of its corners. Large boxes can also be resized from their edges.
* **Layer order:** boxes are layered back to front, so later boxes sit in front of earlier ones. Use the layer buttons to move a box backward or forward.
* **Delete:** tap the bin on the selected box, or the clear-all icon at the top to remove every box.

Tap **Done** to keep your layout. Tap **✕** to leave without saving changes.

> Boxes are stored relative to the canvas. If you change the aspect ratio afterwards, they stretch with it, so set the aspect ratio first.

### 3. Set up each box

Select a box to edit it in the panel below the canvas.

**Type.** Every box is one of three types, each with its own colour on the canvas:

| Type          | Use it for                                | Colour |
| ------------- | ----------------------------------------- | ------ |
| **Object**    | Things in the scene: a cat, a sofa, a cup | Blue   |
| **Character** | A person or creature                      | Orange |
| **Text**      | Words that should appear in the image     | Purple |

Once a box has a palette, it shows the palette's first colour instead.

**Description.** Say what goes in the box. Keep it short and specific.

![A Character box](https://content.gitbook.com/content/tbrc6W2n35oVxftMovDQ/blobs/oVOA1q6OBnZDBhemTwox/04-box-character.png)

**Text** (Text boxes only). Type the exact words to render. Use the description for how they should look, for example "cool font".

![A Text box: the words go in Text, the look goes in Description](https://content.gitbook.com/content/tbrc6W2n35oVxftMovDQ/blobs/s4LCa6lGNku1Xt71YJGU/06-box-text.png)

**Colour palette.** Tap preset colours or the colour wheel to add up to 5 colours. Tap a colour in the bar to remove it. The box takes its first colour on the canvas.

![An Object box with a white and orange palette](https://content.gitbook.com/content/tbrc6W2n35oVxftMovDQ/blobs/W6baddj7Cyz2Qd8ozFqB/05-box-object-palette.png)

## Example

This is the layout from the screenshots above.

**Prompt:** `a man is hugging anime girl in living room. a cat is passing by, realistic style`

**Model:** MOSS Image · **Task:** Text to Image · **Aspect ratio:** 3:2

| # | Type      | Description                                               | Text | Palette       |
| - | --------- | --------------------------------------------------------- | ---- | ------------- |
| 1 | Character | realistic western man, realistic style                    |      |               |
| 2 | Character | anime girl, reze, chainsaw man, sleeveless shirt, smiling |      |               |
| 3 | Object    | cat                                                       |      | White, orange |
| 4 | Text      | cool font                                                 | MOSS | Pink          |
| 5 | Object    | puffy bear                                                |      | Brown         |

**Result**

![The result](https://content.gitbook.com/content/tbrc6W2n35oVxftMovDQ/blobs/Tj7JPvMPiUEm9qhRoz49/07-result.png)

* "MOSS" is written across the top in pink, inside box 4.
* The two characters sit side by side where boxes 1 and 2 were drawn, each in its own style.
* The cat in box 3 follows its white and orange palette.
* The brown bear fills the small box 5 on the left, even though the prompt never mentions it.

## Tips

* **Use the prompt for the whole scene and the boxes for placement.** Describe the setting and mood in the prompt, and what goes where in the boxes.
* **Draw the background first and the foreground last.** Later boxes sit in front, so put the subjects you want in front at the end.
* **Mark people as Character.** This keeps several people in one image distinct from each other.
* **Keep text boxes to a title, a sign or a short phrase.** Put the exact words in Text and the style in Description.
* **Use palettes for key colours,** such as a brand colour or a pet's fur, rather than on every box.
