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

# Music Generation

> Generate music from text prompts using NanoGPT's OpenAI-compatible audio/speech endpoint.

## Overview

NanoGPT supports AI music generation through the same OpenAI-compatible Text-to-Speech endpoint. When you specify a music model, the `input` field is treated as a music prompt (not text to speak) and the API returns an audio file.

## Endpoint

```http theme={null}
POST https://nano-gpt.com/api/v1/audio/speech
```

## Choosing A Music Model

Music model availability changes over time. Discover available audio models via `GET https://nano-gpt.com/api/v1/audio-models` (see [Audio Models](/api-reference/endpoint/audio-models)) and select a model intended for music generation.

## Request Format

```json theme={null}
{
  "model": "YOUR_MUSIC_MODEL_ID",
  "input": "A chill lo-fi hip hop beat with soft piano chords and vinyl crackle, 120 BPM"
}
```

If you are using an OpenAI client that enforces the OpenAI TTS schema (for example the official OpenAI SDK), you may need to include a `voice` field. For music models, `voice` is ignored, so it can be any string (many examples use `"alloy"`).

### Parameters

| Parameter | Type   | Required | Description                                                                                                                  |
| --------- | ------ | -------- | ---------------------------------------------------------------------------------------------------------------------------- |
| `model`   | string | Yes      | A music-capable audio model ID (discover via `GET https://nano-gpt.com/api/v1/audio-models`)                                 |
| `input`   | string | Yes      | A text prompt describing the music you want generated                                                                        |
| `voice`   | string | No       | Ignored for music models. Some OpenAI-compatible clients require this field; if so, pass any string (for example `"alloy"`). |

## Response

The response is an audio file (typically MP3). The `Content-Type` header indicates the format.

```http theme={null}
HTTP/1.1 200 OK
Content-Type: audio/mpeg
```

The response body is raw audio bytes.

## Examples

<CodeGroup>
  ```bash cURL theme={null}
  curl -X POST https://nano-gpt.com/api/v1/audio/speech \
    -H "Authorization: Bearer $NANOGPT_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "YOUR_MUSIC_MODEL_ID",
      "input": "An energetic electronic dance track with heavy bass drops and synth leads"
    }' \
    --output music.mp3
  ```

  ```python Python (OpenAI SDK) theme={null}
  from openai import OpenAI

  client = OpenAI(
      base_url="https://nano-gpt.com/api/v1",
      api_key="YOUR_API_KEY",
  )

  response = client.audio.speech.create(
      model="YOUR_MUSIC_MODEL_ID",
      input="A peaceful acoustic guitar melody with soft drums",
      voice="alloy",  # Ignored for music models; included for OpenAI SDK compatibility
  )

  response.stream_to_file("output.mp3")
  ```
</CodeGroup>

## Tips

* Be descriptive: include genre, instruments, tempo (BPM), mood, and style.
* Duration: generation duration varies by model. Most models produce \~10-30 seconds by default; duration may be influenced by the prompt.
* Cost/quality vary by model. If cost predictability matters, prefer models with flat per-generation pricing (when available).