# AuK (evaluation)

Evaluation only (Qwen Research License): text-to-speech in English and Chinese with a default voice, a voice designed from a description, or a voice cloned from a short clip.

Model id: `auk`
Regions: ATL
Hugging Face: [tencent/AuK](https://huggingface.co/tencent/AuK)

## Pricing

Free

## Limits

Per subscription, on this model:

- Speech: 10 requests a minute, 1 at once, this model's own (the default is 20 and 5)

## Inputs

- text: max prompt 4,800 characters

## Output: speech

| Parameter | Type | Values | Default |
| --- | --- | --- | --- |
| `voice` | string | default, design, clone |  |
| `instructions` | any |  |  |
| `language` | any |  |  |
| `input_references` | array | up to 2 items |  |
| `response_format` | string | mp3, opus, flac, wav, pcm | mp3 |
| `stream_format` | string | audio, sse | audio |
| `speed` | number | 0.25–4 | 1 |
| `seed` | integer |  |  |
| `num_inference_steps` | number | 4–64 | 32 |
| `guidance_scale` | number | 1–5 | 2 |

## Endpoints

- [`POST /v1/audio/speech`](/docs/reference/create-speech.md): Create Speech