# Meta Muse

Meta Muse: plans its layout and checks its own drafts, with built-in world knowledge.

Last updated: 2026-10-05

**Key facts**

- Meta Muse (Muse Image 1.0) is Meta's image model that plans its layout, reviews its own drafts and can look up real places and facts before it draws.
- Meta Muse makes 1K images in Sloosh: you pick the aspect ratio and Meta Muse picks the exact pixel size.
- Meta Muse takes up to 10 reference images in Sloosh and prompts up to 5,000 characters.
- Meta Muse is the only model with Images per run (1 to 10) on Spacelab's Image Model node.

Web page: https://sloosh.ai/docs/meta-muse

Meta Muse thinks before it draws: it plans the layout, reviews its own drafts and improves them. It can look up real places and facts, which helps with concept-heavy scenes.

## When to use it

- Scenes with many specific objects in a particular arrangement.
- Images that depend on real-world knowledge, such as a known landmark or a how-to.
- Short labels and price tags. Long text varies from run to run.

## How to prompt it

- Write it the way you'd ask a person: plain, conversational sentences.
- Be specific about style, lighting and composition.
- Keep text short and state numbers outright. Pair each label with its item, for example "lamp $25".
- With references, describe each one by what it shows. When combining several, expect it to rearrange the layout rather than copy it.
- For a series, repeat the same style words and "same subject" in every prompt.
- You pick the aspect ratio; it picks the exact pixel size.

## Example

Text only: a lifestyle photo with listed contents and a small price tag.

![A woven picnic basket with lemonade, peaches and a book on a striped blanket in a sunny park, with a small tag](https://sloosh.ai/docs/m-meta-muse.webp)

_4:3, 1K, Enhance off. Check short text like the tag on every run: Muse's text varies._

**Prompt**

```text
Make a bright lifestyle photo of a woven picnic basket on a striped blanket in a sunny park. Inside: a glass bottle of lemonade, two peaches and a paperback book. Soft afternoon light, shallow depth of field, warm summer colours. Add a small tag on the basket that reads "PICNIC KIT $45".
```

## Settings in Sloosh

- Aspect ratio: Auto, 21:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16 or 9:21.
- Resolution: 1K.
- References: up to 10 images.
- Prompt: up to 5,000 characters.

Open it in [Image](https://sloosh.ai/studio/image) and pick Meta Muse in the model chip. The Generate button shows the cost first.

## Frequently asked questions

### What is Meta Muse best for?

Meta Muse is best for scenes with many specific objects in a particular arrangement, and for images that depend on real-world knowledge, such as a known landmark or a how-to visual. It also handles short labels and price tags, though longer text varies from run to run.

### Can Meta Muse put text in an image?

Yes, for short text. Keep the words few and state numbers outright, pairing each label with its item, such as "lamp $25". Meta Muse's text varies from run to run, so check every result and run it again if a tag comes out wrong.

### How do I make a consistent series of images with Meta Muse?

Repeat the same style words and a short "same subject" phrase in every Meta Muse prompt, and add the same reference images again on every run. Meta recommends both for keeping a series consistent, since every Meta Muse run varies.

### How should I describe reference images to Meta Muse?

Describe each reference by what it shows, such as "the striped blanket" or "the woman in the red coat". When you combine several references, expect Meta Muse to rearrange the layout and redraw the contents rather than copy your inputs exactly.

### Can Meta Muse make several images in one run?

Yes, in Spacelab. On the Image Model node, Images per run (1 to 10) is offered for Meta Muse only. In Studio, the outputs chip makes up to 4 results with Meta Muse, as with any other model.

## Sources

- [Meta: image generation docs](https://dev.meta.ai/docs/image-generation)
- [Meta: build with Muse Image](https://developer.meta.com/ai/resources/blog/build-with-muse-Image/)