接口简介

本接口适用于 录音文件极速识别场景，基于大模型能力提供识别效果更佳、返回更快的体验。调用形式为 一次请求即返回识别结果，无需 submit/query 轮询。

使用限制

接口地址

POST https://openspeech.bytedance.com/api/v3/auc/bigmodel/recognize/flash

请求 Header

请求体格式（Body）

{
  "user": {
    "uid": "你的AppID"
  },
  "audio": {
    "url": "https://example.com/audio.wav"
    // 或
    "data": "base64编码音频内容"
  },
  "request": {
    "model_name": "bigmodel"
  }
}

注：audio.url 与 audio.data 二选一

请求字段同录音文件标准版本，移除callback与callback_data字段

返回格式（成功）

响应头

服务端返回的 logid，建议用户获取和打印方便定位问题

recognize task response header X-Api-Status-Code: 20000000
recognize task response header X-Api-Message: OK
Thu Jun 19 19:43:55 2025 recognize task response header X-Tt-Logid: 202506191943547B30C313640AF5B35A86

响应体

{
  "audio_info": {
    "duration": 2499
  },
  "result": {
    "additions": {
      "duration": "2499"
    },
    "text": "关闭透传。",
    "utterances": [
      {
        "end_time": 1530,
        "start_time": 450,
        "text": "关闭透传。",
        "words": [
          {
            "confidence": 0,
            "end_time": 770,
            "start_time": 450,
            "text": "关"
          },
          {
            "confidence": 0,
            "end_time": 970,
            "start_time": 770,
            "text": "闭"
          },
          {
            "confidence": 0,
            "end_time": 1210,
            "start_time": 1130,
            "text": "透"
          },
          {
            "confidence": 0,
            "end_time": 1530,
            "start_time": 1490,
            "text": "传"
          }
        ]
      }
    ]
  }
}

错误码

Demo

import json
import time
import uuid
import requests
import base64

# 辅助函数：下载文件
def download_file(file_url):
    response = requests.get(file_url)
    if response.status_code == 200:
        return response.content  # 返回文件内容（二进制）
    else:
        raise Exception(f"下载失败，HTTP状态码: {response.status_code}")

# 辅助函数：将本地文件转换为Base64
def file_to_base64(file_path):
    with open(file_path, 'rb') as file:
        file_data = file.read()  # 读取文件内容
        base64_data = base64.b64encode(file_data).decode('utf-8')  # Base64 编码
    return base64_data

# recognize_task 函数
def recognize_task(file_url=None, file_path=None):
    recognize_url = "https://openspeech.bytedance.com/api/v3/auc/bigmodel/recognize/flash"
    # 填入控制台获取的app id和access token
    appid = "{你的appid}"
    token = "{你的access token}"
    
    headers = {
        "X-Api-App-Key": appid,
        "X-Api-Access-Key": token,
        "X-Api-Resource-Id": "volc.bigasr.auc_turbo", 
        "X-Api-Request-Id": str(uuid.uuid4()),
        "X-Api-Sequence": "-1", 
    }

    # 检查是使用文件URL还是直接上传数据
    audio_data = None
    if file_url:
        audio_data = {"url": file_url}
    elif file_path:
        base64_data = file_to_base64(file_path)  # 转换文件为 Base64
        audio_data = {"data": base64_data}  # 使用Base64编码后的数据

    if not audio_data:
        raise ValueError("必须提供 file_url 或 file_path 其中之一")

    request = {
        "user": {
            "uid": appid
        },
        "audio": audio_data,
        "request": {
            "model_name": "bigmodel",
            # "enable_itn": True,
            # "enable_punc": True,
            # "enable_ddc": True,
            # "enable_speaker_info": False,

        },
    }

    response = requests.post(recognize_url, json=request, headers=headers)
    if 'X-Api-Status-Code' in response.headers:
        print(f'recognize task response header X-Api-Status-Code: {response.headers["X-Api-Status-Code"]}')
        print(f'recognize task response header X-Api-Message: {response.headers["X-Api-Message"]}')
        print(time.asctime() + " recognize task response header X-Tt-Logid: {}".format(response.headers["X-Tt-Logid"]))
        print(f'recognize task response content is: {response.json()}\n')
    else:
        print(f'recognize task failed and the response headers are:: {response.headers}\n')
        exit(1)
    return response

# recognizeMode 不变
def recognizeMode(file_url=None, file_path=None):
    start_time = time.time()
    print(time.asctime() + " START!")
    recognize_response = recognize_task(file_url=file_url, file_path=file_path)
    code = recognize_response.headers['X-Api-Status-Code']
    logid = recognize_response.headers['X-Tt-Logid']
    if code == '20000000':  # task finished
        f = open("result.json", mode='w', encoding='utf-8')
        f.write(json.dumps(recognize_response.json(), indent=4, ensure_ascii=False))
        f.close()
        print(time.asctime() + " SUCCESS! \n")
        print(f"程序运行耗时: {time.time() - start_time:.6f} 秒")
    elif code != '20000001' and code != '20000002':  # task failed
        print(time.asctime() + " FAILED! code: {}, logid: {}".format(code, logid))
        print("headers:")
        # print(query_response.content)

def main(): 
    # 示例：通过 URL 或 文件路径选择传入参数
    file_url = "https://example.mp3"
    file_path = "audio/example.mp3"  # 如果你有本地文件，可以选择这个 
    recognizeMode(file_url=file_url)  # 或者 recognizeMode(file_path=file_path)
    # recognizeMode(file_path=file_path)  # 或者 recognizeMode(file_path=file_path)
 
if __name__ == '__main__': 
    main()