> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.empiriolabs.ai/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.empiriolabs.ai/_mcp/server.

# Stable Audio 2.5

> Up-to-3-minute audio from text with text-to-audio, audio-to-audio, and audio inpainting for music production, sound design, and remixing.

![Stable Audio 2.5](https://media.empiriolabs.ai/model-logos/stable-audio-2-5.png)

[Stability AI](/providers/stability) · Audio Generation

`POST /v1/audio/generations`

Up-to-3-minute audio from text with text-to-audio, audio-to-audio, and audio inpainting for music production, sound design, and remixing.

## At a glance

| Field               | Value                                          |
| ------------------- | ---------------------------------------------- |
| Model id            | `stable-audio-2-5`                             |
| Model release date  | 2025-09-10                                     |
| Input modalities    | Text                                           |
| Output modalities   | Audio                                          |
| Context window      | -                                              |
| Weight precision    | -                                              |
| Features            | music_generation, text_to_audio, sound_effects |
| Native inference    | No                                             |
| New                 | No                                             |
| Supported endpoints | `POST /v1/audio/generations`                   |
| Alternate model ids | `stability-audio-2.5`, `stability/audio-2.5`   |

## Pricing

| Charge     | Spec           | Rate   |
| ---------- | -------------- | ------ |
| Generation | per generation | \$0.68 |

## Example request

```bash
curl https://api.empiriolabs.ai/v1/audio/generations \
  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{"model": "stable-audio-2-5", "prompt": "rain on a tin roof", "duration": 8}'
```

## Parameters

| Parameter       | Type    | Required | Default           | Description                                                                                                                                                          |
| --------------- | ------- | -------- | ----------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `prompt`        | string  | yes      | -                 | What to generate.                                                                                                                                                    |
| `mode`          | enum    | no       | `"text-to-audio"` | audio-inpaint regenerates a \[mask_start, mask_end] window of an existing clip while keeping the rest. · Allowed: `text-to-audio`, `audio-to-audio`, `audio-inpaint` |
| `output_format` | enum    | no       | `"mp3"`           | Audio file format. · Allowed: `mp3`, `wav`                                                                                                                           |
| `duration`      | integer | no       | `190`             | Seconds. Up to 3 minutes 10 seconds. · Range: 1 – 190                                                                                                                |
| `steps`         | integer | no       | `8`               | Diffusion steps. The 2.5 turbo model is tuned for very low step counts. · Range: 4 – 8                                                                               |
| `cfg_scale`     | number  | no       | `1`               | Classifier-free guidance. The turbo model uses small CFG by default. · Range: 1 – 25                                                                                 |
| `strength`      | number  | no       | `1`               | Audio-to-audio only. 0.01 = ignore reference, 1 = stay close to reference. · Range: 0.01 – 1                                                                         |
| `mask_start`    | number  | no       | `30`              | Inpaint window start in seconds, for audio-inpaint. · Range: 0 – 190                                                                                                 |
| `mask_end`      | number  | no       | `190`             | Inpaint window end in seconds, for audio-inpaint. · Range: 0 – 190                                                                                                   |
| `random_seed`   | boolean | no       | true              | If true, use a random seed each call.                                                                                                                                |
| `seed`          | number  | no       | -                 | Reproducibility seed. Used unless random_seed is true.                                                                                                               |
| `audio_url`     | string  | no       | -                 | Reference audio URL for audio-to-audio / inpaint.                                                                                                                    |

## Notes

Adds **audio-inpaint** mode (regenerate a time window) on top of Stable Audio 2.0.

**Mode requirements**

* Audio-to-audio and audio-inpaint both require BOTH a prompt and an uploaded audio file
* Audio-to-audio uses the reference audio for style/conditioning, NOT for voice cloning

---

*Machine-readable schema:* `GET https://api.empiriolabs.ai/v1/models/stable-audio-2-5`.