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

# Generate a world

Describe a place, or show it a photograph, and get a full 3D world back as a Gaussian splat.

Building a landscape by hand means placing terrain, trees, and horizon yourself. Generate World skips that. Describe the place, or hand it a photograph, and it returns a full 3D world you can walk into.

The result arrives as a Gaussian splat, so it carries photographic texture rather than clean geometry. See [Import Gaussian splats](/build/import-gaussian-splats.md) for how splats behave once they are in a scene.

World generation shipped in release 2.0.66.

![The Generate World modal on the Text tab, showing the prompt field, Name, and the Draft, Standard and High detail options](https://1179478100-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSWrjIqlgVAblx6vHGJiM%2Fuploads%2Fgit-blob-e7bc0350bdf37500622c8e65ff7c1c513f3d8051%2Fgenerate-world-text-01.png?alt=media)

## Three ways in

The modal opens on three tabs.

| Tab              | You supply                         |
| ---------------- | ---------------------------------- |
| **Text**         | A written description of the place |
| **Single Image** | One photograph                     |
| **Multi View**   | Several images of the same place   |

### From text

Describe the place in detail. The field takes up to 2000 characters, and the counter tracks what you have used.

Detail earns its length here. Say what the light is doing, what the ground is made of, and where the horizon sits, rather than naming the place and stopping.

### From an image

Upload the photograph and the world is built from it. PNG, JPG and WEBP are accepted.

**Multi View is not a pile of images.** It has named slots: **Front**, which is required, then Back, Left and Right, then up to four more under Additional Images. Eight in total. The modal says why in its own banner: for best results, give it images from multiple sides. A **World description** field sits below the upload and is optional. Use it to state what the image does not show.

![The Generate World modal on the Single Image tab, with an uploaded photograph, a Name, and an optional World description](https://1179478100-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSWrjIqlgVAblx6vHGJiM%2Fuploads%2Fgit-blob-c0e0d5ccb2bf54dbd5601d719ee859f03051d14a%2Fgenerate-world-image-01.png?alt=media)

## Name and detail

**Name** is required on every tab. It is how the world is identified afterwards.

**Detail** offers three levels.

| Level        | The modal calls it | Typical run      | Use it for                                        |
| ------------ | ------------------ | ---------------- | ------------------------------------------------- |
| **Draft**    | Low fidelity       | 2 to 3 minutes   | Checking whether the idea reads before committing |
| **Standard** | High fidelity      | 6 to 7 minutes   | The default                                       |
| **High**     | Best fidelity      | 10 to 12 minutes | The version you intend to shoot                   |

You do not have to remember those numbers: the modal prints the estimate under each level, and the credit cost with it.

The three levels are three different World Labs Marble models rather than one model turned up, which is why the jump from Draft to Standard costs more than the jump from Standard to High in time but not in kind. High's charge is variable rather than fixed, so the same prompt can cost more on one run than another.

![A generated world open in the scene](https://1179478100-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSWrjIqlgVAblx6vHGJiM%2Fuploads%2Fgit-blob-fbf499f6af20ee70a8f532013ac1045200e3130e%2Fgenerate-world-result-01.png?alt=media)

## Where to find it

Open the asset drawer in Build. The generate control sits there alongside the library.

The button reads **Generate World**, and so does the modal it opens. "Generate Scene" survives only as an internal name in the code, so if you find it in an older note or release description, it is the same thing.

Generation is not instant. See the Detail table above for what each level takes.

## Related

* [Import Gaussian splats](/build/import-gaussian-splats.md) for splats you captured yourself
* [Game Mode](/build/game-mode.md) for walking a generated world at eye level


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://help.intangible.ai/build/generate-world.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.