# Generate Videos

Source: https://docs.openwebui.com/features/chat-conversations/image-generation-and-editing/video-generation

Create short videos from a text prompt, or from a starting image, without leaving the chat. The finished video plays in the conversation. Video models take minutes per clip, so video generation is off for regular users until an administrator allows it.

## Set up video generation (admin)

1. Open **Settings** > **Videos** (admin section) and switch **Video Generation** on.
2. Add a connection with the **+** button under **Manage Video Connections**, fill it in and save.
3. Optionally pick a **Default Video Model**. It is used when neither the chat nor the model picks one.
4. Give people access. In **Users** > **Groups** (or the default permissions), turn on the **Video Generation** permission under **Features**. It is off by default, see [Permissions](/features/authentication-access/rbac/permissions). Admins always have it.
5. On each model that should create videos, open **Workspace** > **Models** > **Edit**, then switch on **Video Generation** under **Features** and keep the **Video Generation** category on under **Builtin Tools**. See [Models](/features/workspace/models).

### Video connections

A connection is one provider with a video API. Two kinds work:

- **OpenAI compatible** servers, such as OpenAI itself or LiteLLM.
- **OpenRouter**, which offers Veo, Kling, Seedance and other video models. Open WebUI picks this automatically for `openrouter.ai` URLs; otherwise set **Provider** to **OpenRouter** under **Advanced**.

| Field | What it does |
| --- | --- |
| URL and Auth | The provider's base URL and how to authenticate (API key, session or OAuth forwarding, Entra ID or none), with optional extra Headers. Use Verify Connection to test them. |
| Model IDs | The video models users can pick. Leave it empty to list the provider's video models live. |
| Video Size and Video Length | Defaults sent with each request, for example 1280x720 and 8 seconds. |
| API Version | An API version query for the provider. Not used with OpenRouter. |
| Additional Parameters | Extra JSON passed through to the provider with every request. |
| Access Control | Which users and groups may use the models of this connection. |

Open WebUI waits up to 15 minutes for a video and checks its progress every 5 seconds. Both can be changed with [`VIDEO_GENERATION_TIMEOUT`](/reference/env-configuration#video_generation_timeout) and [`VIDEO_GENERATION_POLL_INTERVAL`](/reference/env-configuration#video_generation_poll_interval).

## Create a video in chat

1. Switch **Video** on in the **+** menu of the message input. Pick the video model there too if the connections offer several.
2. Describe the video and send. Attach an image first to use it as the first frame.
3. The video appears in the reply when the provider has finished it.

Videos are created through a built-in tool, so the chat model needs **Native** [function calling](/features/extensibility/plugin/tools#tool-calling-modes-default-vs-native). Keep your normal chat model selected, as described for [images](/features/chat-conversations/image-generation-and-editing/usage).

If **Video** is missing from the **+** menu, check that video generation is on in **Settings** > **Videos**, that your role has the **Video Generation** permission and that the model has the **Video Generation** feature on.
