将edge-tts部署到cloudflare上，并兼容OpenAI TTS接口

mortimer2024-12-28 18:57

玩配音的基本都知道，微软的edge-tts是好用免费的语音合成利器，唯一缺点是对国内限流越来越严，不过可以通过部署到 cloudflare 来规避，并且还能白嫖 cloudflare的服务器和带宽资源。

先看效果，完成后将有一个配音api接口和一个web配音界面

以下是部署后的js调用代码

js 复制代码 
const requestBody = {
          "model": "tts-1",
          "input": '这是要合成语音的文字',
          "voice": 'zh-CN-XiaoxiaoNeural',
          "response_format": "mp3",
          "speed": 1.0
        };

const response = await fetch('部署到cloudflare后的网址', {
            method: 'POST',
            headers: {
              'Content-Type': 'application/json',
              'Authorization': `Bearer 部署后的key，随意`,
            },
            body: JSON.stringify(requestBody),
});

接下来说说如何部署到 cloudflare 上

登录 cloudflare 创建一个Workers

网址 dash.cloudflare.com/ 如何登录注册不再赘述

登录后，点击左侧 Workers 和 Pages，打开创建页面

继续点击创建

然后在出现的输入框中填写一个英文名称，作为cloudflare赠送的免费子域名头

点击右下角部署后，在新出现的页面中继续点击编辑代码，进入核心阶段，复制代码

然后删掉里面所有的代码，复制下面的代码去替换

ini 复制代码 
// 自定义api key ，用于防止滥用
const API_KEY = '';
const encoder = new TextEncoder();
let expiredAt = null;
let endpoint = null;
let clientId = "";


const TOKEN_REFRESH_BEFORE_EXPIRY = 3 * 60;  
let tokenInfo = {
    endpoint: null,
    token: null,
    expiredAt: null
};

addEventListener("fetch", event => {
    event.respondWith(handleRequest(event.request));
});

async function handleRequest(request) {
    if (request.method === "OPTIONS") {
        return handleOptions(request);
    }
    
  
    const authHeader = request.headers.get("authorization") || request.headers.get("x-api-key");
    const apiKey = authHeader?.startsWith("Bearer ") 
        ? authHeader.slice(7) 
        : null;

    // 只在设置了 API_KEY 的情况下才验证              
    if (API_KEY && apiKey !== API_KEY) {
        return new Response(JSON.stringify({
            error: {
                message: "Invalid API key. Use 'Authorization: Bearer your-api-key' header",
                type: "invalid_request_error",
                param: null,
                code: "invalid_api_key"
            }
        }), {
            status: 401,
            headers: {
                "Content-Type": "application/json",
                ...makeCORSHeaders()
            }
        });
    }

    const requestUrl = new URL(request.url);
    const path = requestUrl.pathname;
    
    if (path === "/v1/audio/speech") {
        try {
            const requestBody = await request.json();
            const { 
                model = "tts-1",
                input,
                voice = "zh-CN-XiaoxiaoNeural",
                response_format = "mp3",
                speed = '1.0',
                volume='0',
                pitch = '0', // 添加 pitch 参数，默认值为 0
                style = "general"//添加style参数，默认值为general
            } = requestBody;

            let rate = parseInt(String( (parseFloat(speed)-1.0)*100) );
            let numVolume = parseInt( String(parseFloat(volume)*100) );
            let numPitch = parseInt(pitch); 
            const response = await getVoice(
                input, 
                voice, 
                rate>=0?`+${rate}%`:`${rate}%`,
                numPitch>=0?`+${numPitch}Hz`:`${numPitch}Hz`,
                numVolume>=0?`+${numVolume}%`:`${numVolume}%`,
                style,
                "audio-24khz-48kbitrate-mono-mp3"
            );

            return response;

        } catch (error) {
            console.error("Error:", error);
            return new Response(JSON.stringify({
                error: {
                    message: error.message,
                    type: "api_error",
                    param: null,
                    code: "edge_tts_error"
                }
            }), {
                status: 500,
                headers: {
                    "Content-Type": "application/json",
                    ...makeCORSHeaders()
                }
            });
        }
    }

    // 默认返回 404
    return new Response("Not Found", { status: 404 });
}

async function handleOptions(request) {
    return new Response(null, {
        status: 204,
        headers: {
            ...makeCORSHeaders(),
            "Access-Control-Allow-Methods": "GET,HEAD,POST,OPTIONS",
            "Access-Control-Allow-Headers": request.headers.get("Access-Control-Request-Headers") || "Authorization"
        }
    });
}

async function getVoice(text, voiceName = "zh-CN-XiaoxiaoNeural", rate = '+0%', pitch = '+0Hz', volume='+0%',style = "general", outputFormat = "audio-24khz-48kbitrate-mono-mp3") {
    try {
        const maxChunkSize = 2000;  
        const chunks = text.trim().split("\n");


        // 获取每个分段的音频
        //const audioChunks = await Promise.all(chunks.map(chunk => getAudioChunk(chunk, voiceName, rate, pitch, volume,style, outputFormat)));
        let audioChunks=[]
        while(chunks.length>0){
            try{
                let audio_chunk= await getAudioChunk(chunks.shift(), voiceName, rate, pitch, volume,style, outputFormat)
                audioChunks.push(audio_chunk)

            }catch(e){
                return new Response(JSON.stringify({
                    error: {
                        message: String(e),
                        type: "api_error",
                        param: `${voiceName}, ${rate}, ${pitch}, ${volume},${style}, ${outputFormat}`,
                        code: "edge_tts_error"
                    }
                }), {
                    status: 500,
                    headers: {
                        "Content-Type": "application/json",
                        ...makeCORSHeaders()
                    }
                });

            }
        }
       

        // 将音频片段拼接起来
        const concatenatedAudio = new Blob(audioChunks, { type: 'audio/mpeg' });
        const response = new Response(concatenatedAudio, {
            headers: {
                "Content-Type": "audio/mpeg",
                ...makeCORSHeaders()
            }
        });

        
        return response;

    } catch (error) {
        console.error("语音合成失败:", error);
        return new Response(JSON.stringify({
            error: {
                message: error,
                type: "api_error",
                param: null,
                code: "edge_tts_error "+voiceName
            }
        }), {
            status: 500,
            headers: {
                "Content-Type": "application/json",
                ...makeCORSHeaders()
            }
        });
    }
}



//获取单个音频数据
async function getAudioChunk(text, voiceName, rate, pitch,volume, style, outputFormat='audio-24khz-48kbitrate-mono-mp3') {
    const endpoint = await getEndpoint();
    const url = `https://${endpoint.r}.tts.speech.microsoft.com/cognitiveservices/v1`;
    let m=text.match(/\[(\d+)\]\s*?$/);
    let slien=0;
    if(m&&m.length==2){
      slien=parseInt(m[1]);
      text=text.replace(m[0],'')

    }
    const response = await fetch(url, {
        method: "POST",
        headers: {
            "Authorization": endpoint.t,
            "Content-Type": "application/ssml+xml",
            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/127.0.0.0 Safari/537.36 Edg/127.0.0.0",
            "X-Microsoft-OutputFormat": outputFormat
        },
        body: getSsml(text, voiceName, rate,pitch,volume, style,slien)
    });

    if (!response.ok) {
        const errorText = await response.text();
        throw new Error(`Edge TTS API error: ${response.status} ${errorText}`);
    }

    return response.blob();

}

function getSsml(text, voiceName, rate, pitch,volume,style,slien=0) {
   let slien_str='';
   if(slien>0){
    slien_str=`<break time="${slien}ms" />`
   }
    return `<speak xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="http://www.w3.org/2001/mstts" version="1.0" xml:lang="zh-CN"> 
                <voice name="${voiceName}"> 
                    <mstts:express-as style="${style}"  styledegree="2.0" role="default" > 
                        <prosody rate="${rate}" pitch="${pitch}" volume="${volume}">${text}</prosody> 
                    </mstts:express-as> 
                    ${slien_str}
                </voice> 
            </speak>`;

}

async function getEndpoint() {
    const now = Date.now() / 1000;
    
    if (tokenInfo.token && tokenInfo.expiredAt && now < tokenInfo.expiredAt - TOKEN_REFRESH_BEFORE_EXPIRY) {
        return tokenInfo.endpoint;
    }

    // 获取新token
    const endpointUrl = "https://dev.microsofttranslator.com/apps/endpoint?api-version=1.0";
    const clientId = crypto.randomUUID().replace(/-/g, "");
    
    try {
        const response = await fetch(endpointUrl, {
            method: "POST",
            headers: {
                "Accept-Language": "zh-Hans",
                "X-ClientVersion": "4.0.530a 5fe1dc6c",
                "X-UserId": "0f04d16a175c411e",
                "X-HomeGeographicRegion": "zh-Hans-CN",
                "X-ClientTraceId": clientId,
                "X-MT-Signature": await sign(endpointUrl),
                "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/127.0.0.0 Safari/537.36 Edg/127.0.0.0",
                "Content-Type": "application/json; charset=utf-8",
                "Content-Length": "0",
                "Accept-Encoding": "gzip"
            }
        });

        if (!response.ok) {
            throw new Error(`获取endpoint失败: ${response.status}`);
        }

        const data = await response.json();
        const jwt = data.t.split(".")[1];
        const decodedJwt = JSON.parse(atob(jwt));
        
        tokenInfo = {
            endpoint: data,
            token: data.t,
            expiredAt: decodedJwt.exp
        };

        return data;

    } catch (error) {
        console.error("获取endpoint失败:", error);
        // 如果有缓存的token，即使过期也尝试使用
        if (tokenInfo.token) {
            console.log("使用过期的缓存token");
            return tokenInfo.endpoint;
        }
        throw error;
    }
}

function addCORSHeaders(response) {
    const newHeaders = new Headers(response.headers);
    for (const [key, value] of Object.entries(makeCORSHeaders())) {
        newHeaders.set(key, value);
    }
    return new Response(response.body, { ...response, headers: newHeaders });
}

function makeCORSHeaders() {
    return {
        "Access-Control-Allow-Origin": "*", 
        "Access-Control-Allow-Methods": "GET,HEAD,POST,OPTIONS",
        "Access-Control-Allow-Headers": "Content-Type, x-api-key",
        "Access-Control-Max-Age": "86400"  
    };
}

async function hmacSha256(key, data) {
    const cryptoKey = await crypto.subtle.importKey(
        "raw",
        key,
        { name: "HMAC", hash: { name: "SHA-256" } },
        false,
        ["sign"]
    );
    const signature = await crypto.subtle.sign("HMAC", cryptoKey, new TextEncoder().encode(data));
    return new Uint8Array(signature);
}

async function base64ToBytes(base64) {
    const binaryString = atob(base64);
    const bytes = new Uint8Array(binaryString.length);
    for (let i = 0; i < binaryString.length; i++) {
        bytes[i] = binaryString.charCodeAt(i);
    }
    return bytes;
}

async function bytesToBase64(bytes) {
    return btoa(String.fromCharCode.apply(null, bytes));
}

function uuid() {
    return crypto.randomUUID().replace(/-/g, "");
}

async function sign(urlStr) {
    const url = urlStr.split("://")[1];
    const encodedUrl = encodeURIComponent(url);
    const uuidStr = uuid();
    const formattedDate = dateFormat();
    const bytesToSign = `MSTranslatorAndroidApp${encodedUrl}${formattedDate}${uuidStr}`.toLowerCase();
    const decode = await base64ToBytes("oik6PdDdMnOXemTbwvMn9de/h9lFnfBaCWbGMMZqqoSaQaqUOqjVGm5NqsmjcBI1x+sS9ugjB55HEJWRiFXYFw==");
    const signData = await hmacSha256(decode, bytesToSign);
    const signBase64 = await bytesToBase64(signData);
    return `MSTranslatorAndroidApp::${signBase64}::${formattedDate}::${uuidStr}`;
}

function dateFormat() {
    const formattedDate = (new Date()).toUTCString().replace(/GMT/, "").trim() + " GMT";
    return formattedDate.toLowerCase();
}

// 添加请求超时控制
async function fetchWithTimeout(url, options, timeout = 30000) {
    const controller = new AbortController();
    const id = setTimeout(() => controller.abort(), timeout);
    
    try {
        const response = await fetch(url, {
            ...options,
            signal: controller.signal
        });
        clearTimeout(id);
        return response;
    } catch (error) {
        clearTimeout(id);
        throw error;
    }
}

特别需要注意的是顶部一行代码，设置 api key ，防止被他人滥用

ini 复制代码 
// 这是 api key，用于验证可用权限
const API_KEY = '';

绑定自己的域名

默认绑定的域名是 https://输入框填写的子域名头.你的账号名.workers.dev/

但不幸的是该域名在国内被墙，想免翻墙使用，你需要绑定一个自己的域名。

  1. 如果你还没有在 cloudflare上添加过自己的域名，可点击右上角添加--现有域，然后输入自己的域名
  1. 如果在cloudflare上已添加过域名，则点击左侧名称返回管理界面，添加自定义域名

点击 设置--域和路由--添加

再点击自定义域，然后填写已添加到 cloudflare 的域名的子域名，例如我的域名 pyvideotrans.com 已添加cloudflare，那么此处我可以填写 ttsapi.pyvideotrans.com

如下图，添加完毕

此处显示你添加的自定义域

使用 openai sdk 测试

这是兼容openai 的接口，可使用openai sdk 直接测试，如下python代码

python 复制代码 
import logging
from openai import OpenAI
import json
import httpx

api_key = 'adgas213423235saeg'  # 替换为你的实际 API key
base_url = 'https://xxx.xxx.com/v1' # 替换为你的自定义域，默认加 /v1


client = OpenAI(
    api_key=api_key,
    base_url=base_url
)



data = {
    'model': 'tts-1',
    'input': '你好啊，亲爱的朋友们',
    'voice': 'zh-CN-YunjianNeural',
    'response_format': 'mp3',
    'speed': 1.0,
}


try:
    response = client.audio.speech.create(
       **data
    )
    with open('./test_openai.mp3', 'wb') as f:
        f.write(response.content)
    print("MP3 file saved successfully to test_openai.mp3")

except Exception as e:
    print(f"An error occurred: {e}")

搭建一个web界面

接口有了，那么如何搭建类似本文章顶部的web页面呢？

打开该项目 github.com/jianchang51... 下载解压，然后将其中的 index.html/output.css/vue.js 3个文件放在服务器目录下，访问 index.html 即可。

注意在 index.html 搜索 https://ttsapi.pyvideotrans.com, 改为你部署在 cloudflare 的自定义域，否则无法使用

参考

  1. edge-tts-openai-cf-worker
  2. edge-tts
相关推荐
程序猿000001号2 小时前
DeepSeek模型：开启人工智能的新篇章
人工智能·deepseek
梦云澜5 小时前
论文阅读(十四)：贝叶斯网络在全基因组DNA甲基化研究中的应用
论文阅读·人工智能·深度学习
忆~遂愿7 小时前
3大关键点教你用Java和Spring Boot快速构建微服务架构：从零开发到高效服务注册与发现的逆袭之路
java·人工智能·spring boot·深度学习·机器学习·spring cloud·eureka
纠结哥_Shrek7 小时前
pytorch逻辑回归实现垃圾邮件检测
人工智能·pytorch·逻辑回归
辞落山7 小时前
自定义数据集，使用 PyTorch 框架实现逻辑回归并保存模型，然后保存模型后再加载模型进行预测
人工智能·pytorch·逻辑回归
天宇琪云8 小时前
关于opencv环境搭建问题：由于找不到opencv_worldXXX.dll,无法执行代码，重新安装程序可能会解决此问题
人工智能·opencv·计算机视觉
大模型之路8 小时前
大模型（LLM）工程师实战之路（含学习路线图、书籍、课程等免费资料推荐）
人工智能·大模型·llm
deephub9 小时前
十大主流联邦学习框架：技术特性、架构分析与对比研究
人工智能·python·深度学习·机器学习·联邦学习
英国翰思教育10 小时前
留学毕业论文如何利用不同问题设计问卷
人工智能·深度学习·学习·算法·学习方法·论文笔记
gaoenyang76052510 小时前
探索高效图像识别：基于OpenCV的形状匹配利器
人工智能·opencv·计算机视觉
热门推荐
01macOS安装Gradle环境02具身智能之RT-H03Dell服务器升级ubuntu 22.04失败解决04AI 绘画Stable Diffusion 研究（十四）SD 图生图+剪映制作人物说话视频05xgboost: Why not implement distributed XGBoost on top of spark06半导体应用系统一些小知识收集(strip&wafer mapping,EAP&scada)07第一届“启航杯”网络安全挑战赛WP08Vue Router 快速入门09密码学原理技术-第六章-introduction to pulibc-key cryptography10论文高级GPT指令推荐