Create Async Speech Transcription Task

Guide API Reference

Pricing Releases

Create a speech transcription task, supporting multiple audio formats, timestamps, and speaker information output.

POST/v1/audio/asr/tasks

Authorization

AuthorizationstringheaderRequired

HTTP: Bearer Auth

Security Scheme Type: http
HTTP Authorization Scheme: Bearer API_key，Used to verify account information, can be viewed in Project Management > API Key .

Request Header

Content-Typeenum<string>Default:application/jsonRequired

The media type of the request body, please set it toapplication/jsonto ensure the correct format of the requested data.

Available options:application/json

Request Bodyapplication/json

file_id long Required

Audio file ID (obtained through the file upload API).

The uploaded async speech recognition file must comply with the following specifications:

Format: mp3, opus, wav, amr, m4a, ogg
Duration: minimum 1 second, maximum 5 hours
Size: no more than 1GB

modelenum<string>Required

Available model code options:u2-asr

format string Required

Audio file type: mp3, opus, wav, amr, m4a, ogg.

sample_rate integer

Audio sample rate, default 16000.

enable_itn boolean

Whether to enable Arabic numeral conversion (e.g., convert "nineteen ninety-seven" to "1997"), default true.

channel integer

Number of audio channels, 1(mono) / 2(stereo), default is 1.

enable_speaker boolean

Whether to enable speaker separation, valid when channel is mono, default false.

speaker_num integer

Number of speakers (valid only when speaker separation is enabled), defaults to automatic recognition.

word_info boolean

Whether to return word-level timestamps, default false.

context string

Context, used to specify contextual information for the model, limited to 500 characters.

hotwords string[ ]

List of hot words, max 200 words, max 5 characters per word.

Response Body Structure

task_idstring

Transcription task ID

base_respobject

Status code and details of this request

base_resp.status_codeinteger

Status code (0=normal; 100001=parameter error; 100101=authentication failed; 100501=triggered RPM rate limit; 100999=internal system error)

base_resp.status_msgstring

Status details

Start Transcribe

curl --request POST \
  --url https://maas-api.hivoice.cn/v1/audio/asr/tasks \
  --header 'Authorization: Bearer <api_key>' \
  --header 'Content-Type: application/json' \
  --data '
  {
    "file_id": 871009177767936,
    "model": "u2-asr",
    "format": "mp3",
    "sample_rate": 16000,
    "enable_itn": true,
    "channel": 1,
    "enable_speaker": false,
    "speaker_num": 2,
    "word_info": true,
    "context": "Context information (max 500 chars)",
    "hotwords": [
        "metaverse",
        "blockchain"
      ]
  }
'

200

{
  "task_id": "task_1234567890abcdef",
  "base_resp": {
    "status_code": 0,
    "status_msg": "success"
  }
}