AzureOpenAI调用Wisper，将.wav音频转为简体中文

最新推荐文章于 2024-10-01 15:46:12 发布

混混子儿

最新推荐文章于 2024-10-01 15:46:12 发布

阅读量116

点赞数 2

分类专栏：大模型文章标签： flask python 后端

本文链接：https://blog.csdn.net/qdmqdtt/article/details/142495423

版权

大模型专栏收录该内容

4 篇文章 0 订阅

订阅专栏

1.下载相关的库

pip install moviepy
pip install openai
pip install pydub

2.输入要调用的大模型url，api_key，大模型版本以及wisper工具

import os
import fnmatch
from moviepy.editor import VideoFileClip
from openai import AzureOpenAI
from pydub import AudioSegment
from pathlib import Path

AZURE_OPENAI_RESOURCE='https://anchat-gpt-north-cenus.openai.azure.com/'
AZURE_OPENAI_KEY='xxxxxxxxxxxx'
AZURE_OPENAI_VERSION='2024-06-01'
AZURE_OPENAI_DEPLOYMENT='whisper'

3.调用AzureOpenAI的wisper大模型，并用prompt微调大模型生成简体中文，不用prompt会生成繁体中文：

# 创建AzureOpenAI大模型实例
client = AzureOpenAI(
    api_version = AZURE_OPENAI_VERSION,
    api_key = AZURE_OPENAI_KEY,
    azure_deployment = AZURE_OPENAI_DEPLOYMENT,
    azure_endpoint = AZURE_OPENAI_RESOURCE,
)

results = []

# slice_results里面放要处理的多个.wav格式的音频文件

for result in slice_results:
    print(result)
    _audio = open(result, "rb") # 以‘rb’格式打开音频文件

    response = client.audio.transcriptions.create(
            model = AZURE_OPENAI_DEPLOYMENT,  # 调用'wisper'大模型
            file = _audio,   # 要处理的.wav音频文件
            language="zh",   # 指定目标语言为简体中文,
            prompt = "生成简体中文文字",   #指定目标语言为简体中文输出
            timeout = 3000
    )  
    print(response.text) # 输出文字结果
    results.append(response.text) # 并存放在results列表里