# Video Generation

BonzAI generates videos locally using the LTX-2 model (GGUF quantized, \~10 GB).

## Endpoint

```
POST http://localhost:65000/video
```

## Modes

* **Text-to-Video**: Generate a video from a text prompt
* **Image-to-Video**: Animate a static image into a video clip

## Request Format

```json
{
  "prompt": "A cat walking in a garden, cinematic lighting",
  "height": 480,
  "width": 768,
  "num_frames": 81,
  "fps": 24,
  "num_inference_steps": 30,
  "guidance_scale": 3.5,
  "image_base64": "..."
}
```

| Parameter             | Default | Description                                       |
| --------------------- | ------- | ------------------------------------------------- |
| `prompt`              | —       | Text description of the video                     |
| `height`              | 480     | Video height in pixels                            |
| `width`               | 768     | Video width in pixels                             |
| `num_frames`          | 81      | Number of frames to generate                      |
| `fps`                 | 24      | Frames per second                                 |
| `num_inference_steps` | 30      | Denoising steps (higher = better quality, slower) |
| `guidance_scale`      | 3.5     | How closely to follow the prompt                  |
| `image_base64`        | —       | Optional base64 image for image-to-video mode     |

## Companion Video Pipeline

In the roleplay system, companions can generate contextual videos:

1. LLM generates the scene description
2. Optionally, an image is generated first as a keyframe
3. LTX-2 generates the video, conditioned on audio if available

## Minting Video NFTs

Generated videos can be minted on Base. Requires **LVL4** (25,000 BONZAI held).