# Models

> A model makes your images, video, audio and text in Endless. How to choose one, what it costs, and what decides which models you can use.

Source: https://endless.okinawa/docs/concepts/models

A **model** produces the output: an image model renders your prompt, a video model makes a clip,
a chat model writes. You pick one from the model picker wherever you generate.

This page doesn't list models, because they are added, repriced and retired without an app
release. The current list is the picker. People who can view workspace governance (admins by
default) also see the full catalog with costs under **Settings → Governance → Models**, which has
a **Media** tab (images, video, audio; charged per run) and a **Chat** tab (the conversation
itself).

## Choosing a model

* **What it makes**: the picker only offers models that fit what you are doing.
* **What it's good at**: models of the same kind differ (text in images, realism, editing). The
  picker groups models by provider and has a search field.
* **What it costs**: media models show a price in credits ("from" when it depends on settings);
  chat models show a level from $ to $$$$. Costs range from a credit or two to several hundred
  per run. Models priced by input size or length show **"Varies"**. The composer shows the exact
  cost before you send. See [Credits and spending](/docs/concepts/credits-and-spending).
* **Its settings**: controls come from the model, so switching model changes them (for example,
  a reference image field appears only on models that accept one).

The picker starts on the workspace default for that kind of output, marked **Default**. Admins
set it per workspace, team or person. See
[Model availability](/docs/governance/model-availability#choosing-the-default-model).

## Why you can't use a model

| What you see                        | Why                                                                                                                                                   |
| ----------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- |
| Dimmed, with a lock and a plan name | The model needs a paid plan. Selecting it opens billing; nothing runs or is charged. A plan never hides a model.                                      |
| Not in the list                     | The workspace turned it off. It disappears from pickers, `@` mentions and what the chat can use, and the block is enforced when the run is requested. |
| A whole feature missing             | Image generation, video generation or the model picker itself can be turned off for the workspace.                                                    |

Restrictions are set at three scopes: **Workspace**, **Team** and **Person**. A narrower scope
can only turn more models off. A model off for the workspace is off for every team and person.
Changing them is a governance permission, held by admins by default. See
[Roles and permissions](/docs/governance/roles-and-permissions).

## Related

* [Assets and visibility](/docs/concepts/assets-and-visibility): everything a model makes becomes an
  asset, private to whoever ran it until shared.