# MiniMax Music 3 (https://docs.fuser.studio/docs/nodes/audio/minimax-music-3)

_Provided by [Fal](https://fal.ai) — [Learn More](https://fal.ai/models/minimax/music-3)._

Model page: https://fuser.studio/models/minimax-music-3

## Inputs [#inputs]

- **Prompt** (`textarea`) — Describe the music style, mood, vocals, instrumentation, tempo, and arrangement.

- **Lyrics** (`textarea`) — Supply the lyrics to sing, using line breaks between lines. Section tags such as [verse], [chorus], or [instrumental] are placed on their own lines.

- **Duration** (`slider`) — Maximum song length in seconds, from 1 to 300. The model can finish earlier; pricing estimates use this upper bound.

   Optional: `true` • Default: `60` • Minimum: `1` • Maximum: `300`

- **Inference Steps** (`slider`) — Flow-matching steps per denoising chunk. More steps can improve quality and take longer.

   Optional: `true` • Default: `30` • Minimum: `1` • Maximum: `100`

- **Guidance** (`slider`) — Control how strongly the generation follows the music description.

   Optional: `true` • Default: `1.7` • Minimum: `0` • Maximum: `20`

- **Seed** (`seed`) — Reuse a seed for reproducible music generation.

   Optional: `true` • Default: `4432`

## Outputs [#outputs]

- **Generated Song** (`audio`) — The generated music audio file.

- **Actual Duration** (`number`) — The generated song length in seconds, which may be shorter than requested.