python音频转文字调用的是百度智能云的接口,因业务需求会涉及比较多数字,所以这里做了数字的处理,可根据自己的需求修改。
from flask import Flask, request, jsonify
import requestsfrom flask_limiter import Limiterapp = Flask(__name__)
limiter = Limiter(app) # 初始化限流扩展API_KEY = "" #百度智能云上获取
SECRET_KEY = ""import re
from cn2an import an2cn, transformdef replace_chinese_numbers(text):# 使用正则表达式匹配句子中的中文数字chinese_numbers = re.findall(r'[零一二三四五六七八九十百千万]+', text)# 遍历匹配到的中文数字,逐一替换为阿拉伯数字for chinese_number in chinese_numbers:arabic_number = transform(chinese_number, 'cn2an')text = text.replace(chinese_number, arabic_number)return text@app.route('/transcribe', methods=['POST'])
@limiter.limit("5 per second") # 设置限流规则为最多同时 5 个请求
def transcribe_audio():audio_data = request.dataaccess_token = get_access_token()if not access_token:return jsonify({"error": "Error getting access token"}), 500url = "https://vop.baidu.com/server_api"headers = {'Content-Type': 'audio/pcm; rate=16000', # 设置正确的 Content-Type'Accept': 'application/json',}params = {"cuid": "your_unique_id", # 替换为你的用户唯一标识,随便写"token": access_token,}response = requests.post(url, headers=headers, params=params, data=audio_data)if response.status_code == 200:try:result = response.json()if "result" in result:transcript = result["result"][0]cleaned_transcript = replace_chinese_numbers(transcript)print(cleaned_transcript)return jsonify({"transcript": cleaned_transcript})else:return jsonify({"error": "No transcription found in the response"}), 500except UserWarning as warning:# 如果出现 UserWarning 异常,返回未处理的 transcriptwarnings.warn(str(warning))return jsonify({"transcript": transcript})else:return jsonify({"error": "Error in transcription request"}), 500def get_access_token():url = "https://aip.baidubce.com/oauth/2.0/token"params = {"grant_type": "client_credentials", "client_id": API_KEY, "client_secret": SECRET_KEY}response = requests.post(url, params=params)if response.status_code == 200:access_token = response.json().get("access_token")return access_tokenelse:print("Error getting access token:", response.text)return Noneif __name__ == '__main__':app.run(host='0.0.0.0', port=16258)