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

# Seed Audio

> Generate voices, music, and sound effects on Mage with Seed Audio: clips of 5 to 120 seconds with audio or image references. Short clips run unlimited from Basic; longer clips need Pro, Pro Plus, or Max.

Mage generates audio as well as images and video. Audio generation runs on Seed Audio, which takes a text prompt plus up to 3 audio references or one image reference and produces clips of 5 to 120 seconds. The unlimited tier follows the clip length: 5 and 10 second clips from Basic, 30 seconds from Pro, 60 seconds from Pro Plus, and 120 seconds on Max.

## Overview

| Detail         |                                  |
| -------------- | -------------------------------- |
| Type           | Audio                            |
| Origin         | Audio model reached through Mage |
| Models         | Seed Audio                       |
| Unlimited from | Basic                            |

## Unlimited generation by plan

Which plan runs each clip length unlimited. The price rises with the duration, and the unlimited tier rises with it.

| Model      | Setting     | Basic     | Pro       | Pro Plus  | Max       |
| ---------- | ----------- | --------- | --------- | --------- | --------- |
| Seed Audio | 5 seconds   | Unlimited | Unlimited | Unlimited | Unlimited |
| Seed Audio | 10 seconds  | Unlimited | Unlimited | Unlimited | Unlimited |
| Seed Audio | 30 seconds  | Gems      | Unlimited | Unlimited | Unlimited |
| Seed Audio | 60 seconds  | Gems      | Gems      | Unlimited | Unlimited |
| Seed Audio | 120 seconds | Gems      | Gems      | Gems      | Unlimited |

Audio and image references do not change the price; only the duration does.

## What it supports on Mage

| Feature          | Support                       |
| ---------------- | ----------------------------- |
| Image references | Up to 1                       |
| Audio references | Up to 3                       |
| LoRAs            | No                            |
| Negative prompts | No                            |
| Durations        | 5, 10, 30, 60, or 120 seconds |

## Best for

* Character dialogue and narration
* Music beds up to 2 minutes
* Sound effects and ambience
* Audio matched to an image you generated

## An example prompt to start from

```text theme={null}
a warm mid-range female narrator reading slowly, small room, no music
```

## Limitations

* 120 seconds is the ceiling for a single generation, and it runs unlimited only on Max. Longer pieces are generated in parts.
* An image reference and audio references are exclusive; a generation uses one kind or the other.
* Several video models generate their own audio, including Plum and MiniMax-H3. If the audio is for a clip on one of those, you may not need a separate generation.