> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zevor.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# AI video

> Describe a scene and AI turns it into up to 8 seconds of cinematic video with native sound and vertical 9:16 format. Lite, Pro and 4K tiers.

Video mode generates vertical 9:16 clips with **native sound**, up to
**8 seconds**, ready to post in under a minute.

## Variants

| Variant             | What you bring            | What it does                                                                       |
| ------------------- | ------------------------- | ---------------------------------------------------------------------------------- |
| **Text to video**   | Just a prompt             | Describe a scene in a sentence; AI turns it into cinematic video. No photo needed. |
| **Image to video**  | 1 image + optional prompt | Upload a photo and AI brings it to life while keeping the style and characters.    |
| **Frames to video** | Start and end image       | You control the first and last frame of the clip.                                  |

## Quality

* **Lite**: cheapest, ideal for iterating to find the prompt.
* **Pro**: publish-ready quality.
* **4K**: max resolution for hero pieces.

<Note>
  If you attach an image, Zevor keeps the quality you picked, it
  doesn't auto-upgrade. You control the cost.
</Note>

## Cost estimate

Credit cost per second of output. The exact total shows on the
**Generate** button before you confirm.

| Quality  | No audio | With audio |
| -------- | -------- | ---------- |
| **Lite** | 4 cr/s   | 6 cr/s     |
| **Pro**  | 6 cr/s   | 8 cr/s     |
| **4K**   | 18 cr/s  | 18 cr/s    |

Example: a Pro video with audio, 8 s = 64 cr. Fits \~2 in a Starter
month (120 cr) or \~9 in Pro (600 cr).

## Prompt examples

* **Text to video**: "Medium cinematic shot, blonde woman in a
  leather jacket walks down a Tokyo street at night, neon reflections
  on the wet pavement, camera follows her from behind."
* **Image to video**: "The model slowly turns to camera and smiles,
  wind moving her hair."
* **Frames to video**: start frame = person from behind; end frame =
  same person looking at camera. AI invents the turn.

## Best practices

* Describe action and camera, not just the subject ("medium shot, she
  turns and looks at camera" beats "a girl").
* For image to video, use a sharp, well-framed image: the video
  inherits its flaws.
* Start with Lite. If you like the result, regenerate in Pro with the
  same prompt for the publishable version.
* If you work with a fixed avatar, see the recipe
  [Build an AI influencer](/en/recipes/ai-influencer).