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

# Music generator

> Generate songs from lyrics and style prompts in the Unsora dashboard

The **Music generator** turns lyrics and an optional style prompt into full songs powered by [Mureka AI](https://wavespeed.ai/models/mureka-ai). Open it from the dashboard sidebar at [`/music-generator`](https://app.tryunsora.com/music-generator).

## Layout

The page has two panels:

1. **Song settings** (left on desktop) — model, lyrics, style prompt, and **Generate song**
2. **Results** (right) — compact cards for in-progress and completed tracks

On mobile, settings stack above results.

## Choose a model first

The **Model** selector is at the top of the song settings panel. Pick the engine before you write lyrics.

| Model                          | Best for                                                                                  |
| ------------------------------ | ----------------------------------------------------------------------------------------- |
| **Mureka V9 (Auto)**           | Default full song with vocals                                                             |
| **Mureka V9 / V8 / O2 / V7.6** | Full songs with lyrics; credit cost varies by model                                       |
| **Mureka V7.5 (BGM)**          | Instrumental background music — style prompt is required unless lyrics describe mood only |

<Note>
  Song models require **Lyrics**. BGM mode treats lyrics as optional and focuses on the style prompt.
</Note>

## Write lyrics and style

**Lyrics** — paste or write verse/chorus text. Section labels like `[Verse]` and `[Chorus]` are supported.

**Style prompt** (optional for songs, important for BGM) — plain language for genre, tempo, mood, and vocal style, for example: `r&b, slow, passionate, male vocal`.

Click **Generate song** (or **Generate BGM** for V7.5). Credits are deducted when the job is queued.

## Results cards

Each result is a compact dashboard row:

* **Play / pause** on the artwork button
* **Title** from your style prompt (or “Untitled song”)
* **One-line lyric preview** when idle
* **Rolling synced lyrics** while that track is the active player (see [Playback & player](/guides/music-generator/playback))
* **Download** and **delete** on hover (always visible on touch devices)

In-progress jobs show a spinner and status (`Queued`, `Generating…`). Failed jobs show the error message and can be dismissed.

## Tips

* Keep lyrics under 3000 characters (API limit).
* Use a style prompt even when optional — it steers genre and vocal delivery.
* You can start playback and switch to other tools; audio keeps playing in the global player.

## API access

Automate the same workflow with the public API:

* [Create song](/api-reference/music-generations/create)
* [Poll music status](/api-reference/music/poll-status)
* [List generations](/api-reference/music-generations/list)