> 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.

# StepAudio 3 ASR Max

> Context-aware transcription for proper nouns, technical terms, dialects, and difficult audio including whispers, fast speech, and singing.

![StepAudio 3 ASR Max](https://media.empiriolabs.ai/model-logos/stepaudio-3-asr-max.png)

[StepFun](/providers/stepfun) · Transcription

`POST /v1/audio/transcriptions`

Context-aware transcription for proper nouns, technical terms, dialects, and difficult audio including whispers, fast speech, and singing.

## At a glance

| Field               | Value                                       |
| ------------------- | ------------------------------------------- |
| Model id            | `stepaudio-3-asr-max`                       |
| Model release date  | 2026-08-31                                  |
| Input modalities    | Audio                                       |
| Output modalities   | Text                                        |
| Context window      | -                                           |
| Weight precision    | -                                           |
| Region              | International                               |
| Features            | transcription, speech_to_text, multilingual |
| Native inference    | No                                          |
| New                 | Yes                                         |
| Supported endpoints | `POST /v1/audio/transcriptions`             |
| Alternate model ids | `stepfun/stepaudio-3-asr-max`               |

## Pricing

| Charge        | Spec              | Rate   |
| ------------- | ----------------- | ------ |
| Transcription | per hour of audio | \$0.24 |

## Example request

```bash
curl https://api.empiriolabs.ai/v1/audio/transcriptions \
  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
  -F model=stepaudio-3-asr-max \
  -F file=@meeting.mp3
```

## Parameters

| Parameter      | Type    | Required | Default | Description                                                                                         |
| -------------- | ------- | -------- | ------- | --------------------------------------------------------------------------------------------------- |
| `file`         | string  | no       | -       | Audio file upload for transcription.                                                                |
| `file_url`     | string  | no       | -       | Public URL to an audio file.                                                                        |
| `audio_base64` | string  | no       | -       | Base64 audio payload for JSON requests.                                                             |
| `enable_itn`   | boolean | no       | true    | Enable inverse text normalization, which writes numbers, dates, and currency in their written form. |
| `format`       | enum    | no       | `"wav"` | Audio container format. · Allowed: `wav`, `mp3`, `ogg`, `m4a`, `pcm`                                |
| `codec`        | string  | no       | -       | PCM codec such as pcm_s16le.                                                                        |
| `rate`         | number  | no       | -       | PCM sample rate in Hz.                                                                              |
| `bits`         | number  | no       | -       | PCM bit depth.                                                                                      |
| `channel`      | number  | no       | -       | PCM channel count.                                                                                  |

## Notes

Supports wav, mp3, ogg, m4a, and pcm input. PCM requests should include codec, sample rate, bit depth, and channel count. The language is detected automatically and cannot be set; Chinese, English, Japanese, Korean, French, and Spanish are recognized, with languages other than Chinese and English in preview. Inverse text normalization is on by default and can be turned off with enable_itn. Hotwords and word timestamps are not available on this model. Audio is billed by its measured duration, with a one second minimum.

---

*Machine-readable schema:* `GET https://api.empiriolabs.ai/v1/models/stepaudio-3-asr-max`.