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

# Fish Audio Drama 3

> Expressive speech you direct with inline cues

> [card] [Try Fish Audio Drama 3 in the Workbench](https://www.oxen.ai/ai/workbench?model=fish-audio-drama-3)
>
> Run this model interactively, tune parameters, and compare outputs.

**Model ID:** `fish-audio-drama-3`

Fish Audio Drama 3 is a text to speech model built for performances you direct from inside the script, across more than 80 languages.

Directions in square brackets, such as `[barely holding it together]` or `[sighs]`, are acted out where they're written, and tag pairs like `&lt;whisper>…&lt;/whisper>` change how the words inside them are said. It suits character scenes, audio drama, and long-form narration. Language tags such as `&lt;french>…&lt;/french>` switch the language for a phrase, though they're still experimental. The weights are published under the Fish Audio Community License.

| Metric | Value |
| - | - |
| Parameter Count | 35B (3B active) |
| Mixture of Experts | Yes |
| Multilingual | Yes, 80+ languages |

## Example request

> [!TIP]
>
> Use the [Workbench](https://www.oxen.ai/ai/workbench?model=fish-audio-drama-3) as a request builder: configure parameters for this model in the UI, then open the **API** tab to copy the exact cURL or Python call.

> [tab] Minimal
>
> ```bash cURL theme={null}
> curl -X POST https://hub.oxen.ai/api/ai/audio/speech \
>   -H "Content-Type: application/json" \
>   -H "Authorization: Bearer $OXEN_API_KEY" \
>   -d '{
>   "model": "fish-audio-drama-3",
>   "prompt": "<prompt>",
>   "input": "Welcome to Oxen"
> }'
> ```
>
> ```python Python theme={null}
> import os
> import requests
>
> response = requests.post(
>     "https://hub.oxen.ai/api/ai/audio/speech",
>     headers={
>         "Content-Type": "application/json",
>         "Authorization": f"Bearer {os.environ['OXEN_API_KEY']}",
>     },
>     json={
>         "model": "fish-audio-drama-3",
>         "prompt": "<prompt>",
>         "input": "Welcome to Oxen"
>     },
> )
> response.raise_for_status()
> print(response.json())
> ```

> [tab] All parameters
>
> ```bash cURL theme={null}
> curl -X POST https://hub.oxen.ai/api/ai/audio/speech \
>   -H "Content-Type: application/json" \
>   -H "Authorization: Bearer $OXEN_API_KEY" \
>   -d '{
>   "model": "fish-audio-drama-3",
>   "prompt": "<prompt>",
>   "voice": "933563129e564b19a115bedd57b7406a",
>   "output_format": "mp3",
>   "speed": 1.0,
>   "input": "Welcome to Oxen"
> }'
> ```
>
> ```python Python theme={null}
> import os
> import requests
>
> response = requests.post(
>     "https://hub.oxen.ai/api/ai/audio/speech",
>     headers={
>         "Content-Type": "application/json",
>         "Authorization": f"Bearer {os.environ['OXEN_API_KEY']}",
>     },
>     json={
>         "model": "fish-audio-drama-3",
>         "prompt": "<prompt>",
>         "voice": "933563129e564b19a115bedd57b7406a",
>         "output_format": "mp3",
>         "speed": 1.0,
>         "input": "Welcome to Oxen"
>     },
> )
> response.raise_for_status()
> print(response.json())
> ```

## Fetch model details

The [models endpoint](https://docs.oxen.ai/inference-api/reference/models/overview.md) returns the full model object, including its `json_request_schema`.

```bash theme={null}
curl -H "Authorization: Bearer $OXEN_API_KEY" https://hub.oxen.ai/api/ai/models/fish-audio-drama-3
```

## Request parameters

### Required parameters

| Field | Type | Default | Description |
| - | - | - | - |
| `prompt` | `string` | — | Text to speak. Directions in square brackets, like \[nervous] or \[sighs], are acted out instead of read aloud. A tag pair like \<whisper>go on\</whisper> changes how the words inside it are said. |

### Optional parameters

| Field | Type | Default | Description |
| - | - | - | - |
| `voice` | `string` | `"933563129e564b19a115bedd57b7406a"` | The voice that speaks the text. Each voice keeps its own accent in every language it speaks. One of: 933563129e564b19a115bedd57b7406a, bf322df2096a46f18c579d0baa36f41d, b347db033a6549378b48d00acb0d06cd, 536d3a5e000945adb7038665781a4aca, 9a9cf47702da476aa4629e2506d4a857, 79d0bd3e4e5444b18f7b6d89b5927bf1, e3cd384158934cc9a01029cd7d278634, 0db6e93af420470b9d9b9a886dbaa954, f3a2b90078d54a65af4b95f63bc7798e, c70cfb98867a4bfb9ae20d2401257b92, 74c6aba5cbf94a15bbdc547ffce5cb38, 5d29a99739c14d4ca3e4fe42193105b2, 97eda32449cf443bbc4b8782853ca554, 297a6fd278df47c3b9da9bfdf55ac89a, 96ed4e3f47334430b990df471d0a0268, 1faa840246094164bf4b0b0b94ae1675, 3f065d660212478a9c697fa965b4d098, ec0718028a2445fb9be8e8bfb56a0150, 5f2350f72d864ba0915668843dd1a584, 0ffee1af32254090b3630823baade12b, 10a3a20742114a4ea6dd441e7591850f, f69bca092b674168a8d02d61ca20943c, 3937b44b604249f18b94913daef94043, 7b7c88bf825a4ab3bca27d55c26ddf16, ef02f46e6a7540759f9d767d7e22c2e5, 59ac2b0ad66641b5ab4666d3ed89e5d4, bf75c2e80b0e441485eeec20330a75fa, b871e425b6c0442baa5cf73f6c336bbe, 0b14f34a13a94ed88fe6113193749bb0, b26519ebdcd14bec967c020c5aabc927, 2f64e528a4f2406691925ab1e0ed5b7f, b080019dc38545f39def724726386d30. |
| `output_format` | `string` | `"mp3"` | Audio file format. One of: mp3, wav. |
| `speed` | `number` | `1.0` | Speaking rate, where 1.0 is the voice's natural pace. Range: 0.5 – 2.0. |


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.