Spaces:
Running
Running
Update main.ts
Browse files
main.ts
CHANGED
|
@@ -1,154 +1,449 @@
|
|
| 1 |
-
import { serve } from "https://deno.land/std/http/server.ts";
|
| 2 |
-
import { EdgeSpeechTTS } from "https://esm.sh/@lobehub/tts@1";
|
| 3 |
-
|
| 4 |
-
const AUTH_TOKEN = Deno.env.get("AUTH_TOKEN");
|
| 5 |
-
const VOICES_URL = "https://speech.platform.bing.com/consumer/speech/synthesize/readaloud/voices/list?trustedclienttoken=6A5AA1D4EAFF4E9FB37E23D68491D6F4";
|
| 6 |
-
|
| 7 |
-
async function fetchVoiceList() {
|
| 8 |
-
const response = await fetch(VOICES_URL);
|
| 9 |
-
const voices = await response.json();
|
| 10 |
-
return voices.reduce((acc: Record<string, { model: string, name: string, friendlyName: string, locale: string }[]>, voice: any) => {
|
| 11 |
-
const { ShortName: model, ShortName: name, FriendlyName: friendlyName, Locale: locale } = voice;
|
| 12 |
-
if (!acc[locale]) acc[locale] = [];
|
| 13 |
-
acc[locale].push({ model, name, friendlyName, locale });
|
| 14 |
-
return acc;
|
| 15 |
-
}, {});
|
| 16 |
-
}
|
| 17 |
-
|
| 18 |
-
async function synthesizeSpeech(model: string, voice: string, text: string) {
|
| 19 |
-
let voiceName;
|
| 20 |
-
let rate = 0;
|
| 21 |
-
let pitch = 0;
|
| 22 |
-
|
| 23 |
-
if (model.includes("tts")) {
|
| 24 |
-
rate = 0.1;
|
| 25 |
-
pitch = 0.2;
|
| 26 |
-
|
| 27 |
-
switch (voice) {
|
| 28 |
-
case "alloy":
|
| 29 |
-
voiceName = "zh-CN-YunjianNeural";
|
| 30 |
-
break;
|
| 31 |
-
case "echo":
|
| 32 |
-
voiceName = "zh-CN-YunyangNeural";
|
| 33 |
-
break;
|
| 34 |
-
case "fable":
|
| 35 |
-
voiceName = "zh-CN-XiaoxiaoNeural";
|
| 36 |
-
break;
|
| 37 |
-
default:
|
| 38 |
-
voiceName = "zh-CN-YunxiNeural";
|
| 39 |
-
break;
|
| 40 |
-
}
|
| 41 |
-
} else {
|
| 42 |
-
voiceName = model;
|
| 43 |
-
const params = Object.fromEntries(
|
| 44 |
-
voice.split("|").map((p) => p.split(":") as [string, string])
|
| 45 |
-
);
|
| 46 |
-
rate = Number(params["rate"] || 0);
|
| 47 |
-
pitch = Number(params["pitch"] || 0);
|
| 48 |
-
}
|
| 49 |
-
|
| 50 |
-
const tts = new EdgeSpeechTTS();
|
| 51 |
-
|
| 52 |
-
const payload = {
|
| 53 |
-
input: text,
|
| 54 |
-
options: {
|
| 55 |
-
rate: rate,
|
| 56 |
-
pitch: pitch,
|
| 57 |
-
voice: voiceName
|
| 58 |
-
},
|
| 59 |
-
};
|
| 60 |
-
const response = await tts.create(payload);
|
| 61 |
-
const mp3Buffer = new Uint8Array(await response.arrayBuffer());
|
| 62 |
-
|
| 63 |
-
console.log(`Successfully synthesized speech, returning audio/mpeg response`);
|
| 64 |
-
return new Response(mp3Buffer, {
|
| 65 |
-
headers: { "Content-Type": "audio/mpeg" },
|
| 66 |
-
});
|
| 67 |
-
}
|
| 68 |
-
|
| 69 |
-
function
|
| 70 |
-
const
|
| 71 |
-
|
| 72 |
-
}
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
}
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154 |
});
|
|
|
|
| 1 |
+
import { serve } from "https://deno.land/std/http/server.ts";
|
| 2 |
+
import { EdgeSpeechTTS } from "https://esm.sh/@lobehub/tts@1";
|
| 3 |
+
|
| 4 |
+
const AUTH_TOKEN = Deno.env.get("AUTH_TOKEN");
|
| 5 |
+
const VOICES_URL = "https://speech.platform.bing.com/consumer/speech/synthesize/readaloud/voices/list?trustedclienttoken=6A5AA1D4EAFF4E9FB37E23D68491D6F4";
|
| 6 |
+
|
| 7 |
+
async function fetchVoiceList() {
|
| 8 |
+
const response = await fetch(VOICES_URL);
|
| 9 |
+
const voices = await response.json();
|
| 10 |
+
return voices.reduce((acc: Record<string, { model: string, name: string, friendlyName: string, locale: string }[]>, voice: any) => {
|
| 11 |
+
const { ShortName: model, ShortName: name, FriendlyName: friendlyName, Locale: locale } = voice;
|
| 12 |
+
if (!acc[locale]) acc[locale] = [];
|
| 13 |
+
acc[locale].push({ model, name, friendlyName, locale });
|
| 14 |
+
return acc;
|
| 15 |
+
}, {});
|
| 16 |
+
}
|
| 17 |
+
|
| 18 |
+
async function synthesizeSpeech(model: string, voice: string, text: string) {
|
| 19 |
+
let voiceName;
|
| 20 |
+
let rate = 0;
|
| 21 |
+
let pitch = 0;
|
| 22 |
+
|
| 23 |
+
if (model.includes("tts")) {
|
| 24 |
+
rate = 0.1;
|
| 25 |
+
pitch = 0.2;
|
| 26 |
+
|
| 27 |
+
switch (voice) {
|
| 28 |
+
case "alloy":
|
| 29 |
+
voiceName = "zh-CN-YunjianNeural";
|
| 30 |
+
break;
|
| 31 |
+
case "echo":
|
| 32 |
+
voiceName = "zh-CN-YunyangNeural";
|
| 33 |
+
break;
|
| 34 |
+
case "fable":
|
| 35 |
+
voiceName = "zh-CN-XiaoxiaoNeural";
|
| 36 |
+
break;
|
| 37 |
+
default:
|
| 38 |
+
voiceName = "zh-CN-YunxiNeural";
|
| 39 |
+
break;
|
| 40 |
+
}
|
| 41 |
+
} else {
|
| 42 |
+
voiceName = model;
|
| 43 |
+
const params = Object.fromEntries(
|
| 44 |
+
voice.split("|").map((p) => p.split(":") as [string, string])
|
| 45 |
+
);
|
| 46 |
+
rate = Number(params["rate"] || 0);
|
| 47 |
+
pitch = Number(params["pitch"] || 0);
|
| 48 |
+
}
|
| 49 |
+
|
| 50 |
+
const tts = new EdgeSpeechTTS();
|
| 51 |
+
|
| 52 |
+
const payload = {
|
| 53 |
+
input: text,
|
| 54 |
+
options: {
|
| 55 |
+
rate: rate,
|
| 56 |
+
pitch: pitch,
|
| 57 |
+
voice: voiceName
|
| 58 |
+
},
|
| 59 |
+
};
|
| 60 |
+
const response = await tts.create(payload);
|
| 61 |
+
const mp3Buffer = new Uint8Array(await response.arrayBuffer());
|
| 62 |
+
|
| 63 |
+
console.log(`Successfully synthesized speech, returning audio/mpeg response`);
|
| 64 |
+
return new Response(mp3Buffer, {
|
| 65 |
+
headers: { "Content-Type": "audio/mpeg" },
|
| 66 |
+
});
|
| 67 |
+
}
|
| 68 |
+
|
| 69 |
+
function validateContentType(req: Request, expected: string) {
|
| 70 |
+
const contentType = req.headers.get("Content-Type");
|
| 71 |
+
if (contentType !== expected) {
|
| 72 |
+
console.log(`Invalid Content-Type ${contentType}, expected ${expected}`);
|
| 73 |
+
return new Response("Bad Request", { status: 400 });
|
| 74 |
+
}
|
| 75 |
+
}
|
| 76 |
+
|
| 77 |
+
async function handleDebugRequest(req: Request) {
|
| 78 |
+
const url = new URL(req.url);
|
| 79 |
+
const voice = url.searchParams.get("voice") || "";
|
| 80 |
+
const model = url.searchParams.get("model") || "";
|
| 81 |
+
const text = url.searchParams.get("text") || "";
|
| 82 |
+
|
| 83 |
+
console.log(`Debug request with model=${model}, voice=${voice}, text=${text}`);
|
| 84 |
+
|
| 85 |
+
if (!voice || !model || !text) {
|
| 86 |
+
console.log("Missing required parameters");
|
| 87 |
+
return new Response("Bad Request", { status: 400 });
|
| 88 |
+
}
|
| 89 |
+
|
| 90 |
+
return synthesizeSpeech(model, voice, text);
|
| 91 |
+
}
|
| 92 |
+
|
| 93 |
+
async function handleSynthesisRequest(req: Request) {
|
| 94 |
+
|
| 95 |
+
if (req.method !== "POST") {
|
| 96 |
+
console.log(`Invalid method ${req.method}, expected POST`);
|
| 97 |
+
return new Response("Method Not Allowed", { status: 405 });
|
| 98 |
+
}
|
| 99 |
+
|
| 100 |
+
const invalidContentType = validateContentType(req, "application/json");
|
| 101 |
+
if (invalidContentType) return invalidContentType;
|
| 102 |
+
|
| 103 |
+
const { model, input, voice } = await req.json();
|
| 104 |
+
console.log(`Synthesis request with model=${model}, input=${input}, voice=${voice}`);
|
| 105 |
+
|
| 106 |
+
return synthesizeSpeech(model, voice, input);
|
| 107 |
+
}
|
| 108 |
+
|
| 109 |
+
|
| 110 |
+
async function handleDemoRequest(req: Request) {
|
| 111 |
+
const groupedVoiceList = await fetchVoiceList();
|
| 112 |
+
|
| 113 |
+
const html = `<!DOCTYPE html>
|
| 114 |
+
<html lang="en">
|
| 115 |
+
<head>
|
| 116 |
+
<meta charset="UTF-8" />
|
| 117 |
+
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
|
| 118 |
+
<title>tts</title>
|
| 119 |
+
<link
|
| 120 |
+
href="https://fonts.googleapis.com/css2?family=Noto+Sans+SC:wght@400;700&display=swap"
|
| 121 |
+
rel="stylesheet"
|
| 122 |
+
/>
|
| 123 |
+
<style>
|
| 124 |
+
:root {
|
| 125 |
+
--primary-color: #6c8bd6;
|
| 126 |
+
--primary-light: #a2b3e3;
|
| 127 |
+
--primary-dark: #3d5b8f;
|
| 128 |
+
--secondary-color: #f08080;
|
| 129 |
+
--text-color: #333;
|
| 130 |
+
--text-secondary: #777;
|
| 131 |
+
--bg-color: #fff;
|
| 132 |
+
}
|
| 133 |
+
body {
|
| 134 |
+
font-family: "Noto Sans SC", "Arial", sans-serif;
|
| 135 |
+
color: var(--text-color);
|
| 136 |
+
margin: 0;
|
| 137 |
+
padding: 0;
|
| 138 |
+
display: flex;
|
| 139 |
+
justify-content: center;
|
| 140 |
+
background-color: #fafafa;
|
| 141 |
+
background-image: linear-gradient(135deg, #f5f7fa 0%, #c3cfe2 100%);
|
| 142 |
+
position: relative;
|
| 143 |
+
overflow: hidden;
|
| 144 |
+
}
|
| 145 |
+
body::before {
|
| 146 |
+
content: "";
|
| 147 |
+
position: absolute;
|
| 148 |
+
top: 0;
|
| 149 |
+
left: 0;
|
| 150 |
+
right: 0;
|
| 151 |
+
bottom: 0;
|
| 152 |
+
background: repeating-radial-gradient(
|
| 153 |
+
circle at 50% 50%,
|
| 154 |
+
rgba(255, 255, 255, 0.8) 0%,
|
| 155 |
+
rgba(255, 255, 255, 0.8) 2%,
|
| 156 |
+
transparent 2%,
|
| 157 |
+
transparent 4%,
|
| 158 |
+
rgba(255, 255, 255, 0.8) 4%,
|
| 159 |
+
rgba(255, 255, 255, 0.8) 6%,
|
| 160 |
+
transparent 6%,
|
| 161 |
+
transparent 8%,
|
| 162 |
+
rgba(255, 255, 255, 0.8) 8%,
|
| 163 |
+
rgba(255, 255, 255, 0.8) 10%,
|
| 164 |
+
transparent 10%
|
| 165 |
+
),
|
| 166 |
+
repeating-linear-gradient(
|
| 167 |
+
45deg,
|
| 168 |
+
#d4f4ff 0%,
|
| 169 |
+
#d4f4ff 5%,
|
| 170 |
+
#e6f9ff 5%,
|
| 171 |
+
#e6f9ff 10%,
|
| 172 |
+
#f0faff 10%,
|
| 173 |
+
#f0faff 15%,
|
| 174 |
+
#e6f9ff 15%,
|
| 175 |
+
#e6f9ff 20%,
|
| 176 |
+
#d4f4ff 20%,
|
| 177 |
+
#d4f4ff 25%
|
| 178 |
+
);
|
| 179 |
+
background-blend-mode: multiply;
|
| 180 |
+
opacity: 0.8;
|
| 181 |
+
z-index: -1;
|
| 182 |
+
animation: glitch 15s infinite;
|
| 183 |
+
}
|
| 184 |
+
.container {
|
| 185 |
+
display: flex;
|
| 186 |
+
max-width: 1200px;
|
| 187 |
+
width: 100%;
|
| 188 |
+
margin: 40px;
|
| 189 |
+
background: #fff;
|
| 190 |
+
border-radius: 12px;
|
| 191 |
+
position: relative;
|
| 192 |
+
background-color: rgba(255, 255, 255, 0.8);
|
| 193 |
+
z-index: 1;
|
| 194 |
+
}
|
| 195 |
+
@keyframes glitch {
|
| 196 |
+
0% {
|
| 197 |
+
background-position: 0 0, 0 0;
|
| 198 |
+
filter: hue-rotate(0deg);
|
| 199 |
+
}
|
| 200 |
+
50% {
|
| 201 |
+
background-position: 10px 10px, -10px 10px;
|
| 202 |
+
filter: hue-rotate(360deg);
|
| 203 |
+
}
|
| 204 |
+
100% {
|
| 205 |
+
background-position: 0 0, 0 0;
|
| 206 |
+
filter: hue-rotate(0deg);
|
| 207 |
+
}
|
| 208 |
+
}
|
| 209 |
+
.input-area,
|
| 210 |
+
.output-area {
|
| 211 |
+
padding: 30px;
|
| 212 |
+
width: 50%;
|
| 213 |
+
}
|
| 214 |
+
.input-area {
|
| 215 |
+
border-right: 1px solid #e0e0e0;
|
| 216 |
+
}
|
| 217 |
+
h1 {
|
| 218 |
+
font-size: 36px;
|
| 219 |
+
color: var(--primary-color);
|
| 220 |
+
margin-bottom: 30px;
|
| 221 |
+
}
|
| 222 |
+
.filter-section {
|
| 223 |
+
margin-bottom: 30px;
|
| 224 |
+
}
|
| 225 |
+
.filter-section label {
|
| 226 |
+
display: block;
|
| 227 |
+
font-size: 16px;
|
| 228 |
+
color: var(--text-secondary);
|
| 229 |
+
margin-bottom: 10px;
|
| 230 |
+
}
|
| 231 |
+
.filter-section input {
|
| 232 |
+
font-size: 16px;
|
| 233 |
+
padding: 10px 15px;
|
| 234 |
+
border: 2px solid var(--primary-light);
|
| 235 |
+
border-radius: 8px;
|
| 236 |
+
outline: none;
|
| 237 |
+
transition: border-color 0.3s, box-shadow 0.3s;
|
| 238 |
+
width: 100%;
|
| 239 |
+
box-sizing: border-box;
|
| 240 |
+
}
|
| 241 |
+
.filter-section input:focus {
|
| 242 |
+
border-color: var(--primary-color);
|
| 243 |
+
box-shadow: 0 0 0 2px var(--primary-light);
|
| 244 |
+
}
|
| 245 |
+
.slider-container {
|
| 246 |
+
margin-bottom: 30px;
|
| 247 |
+
}
|
| 248 |
+
.slider-container label {
|
| 249 |
+
display: block;
|
| 250 |
+
font-size: 16px;
|
| 251 |
+
color: var(--text-secondary);
|
| 252 |
+
margin-bottom: 10px;
|
| 253 |
+
}
|
| 254 |
+
.slider {
|
| 255 |
+
-webkit-appearance: none;
|
| 256 |
+
width: 100%;
|
| 257 |
+
height: 10px;
|
| 258 |
+
border-radius: 5px;
|
| 259 |
+
background: linear-gradient(
|
| 260 |
+
to right,
|
| 261 |
+
var(--secondary-color) 0%,
|
| 262 |
+
var(--primary-color) 50%,
|
| 263 |
+
var(--primary-light) 100%
|
| 264 |
+
);
|
| 265 |
+
box-shadow: inset 0 1px 2px rgba(0, 0, 0, 0.1),
|
| 266 |
+
0 1px rgba(255, 255, 255, 0.1);
|
| 267 |
+
outline: none;
|
| 268 |
+
opacity: 0.7;
|
| 269 |
+
-webkit-transition: 0.2s;
|
| 270 |
+
transition: opacity 0.2s;
|
| 271 |
+
margin-bottom: 10px;
|
| 272 |
+
}
|
| 273 |
+
.slider:hover {
|
| 274 |
+
opacity: 1;
|
| 275 |
+
}
|
| 276 |
+
.slider::-webkit-slider-thumb {
|
| 277 |
+
-webkit-appearance: none;
|
| 278 |
+
appearance: none;
|
| 279 |
+
width: 20px;
|
| 280 |
+
height: 20px;
|
| 281 |
+
border-radius: 50%;
|
| 282 |
+
background: #fff;
|
| 283 |
+
border: 2px solid var(--primary-color);
|
| 284 |
+
cursor: pointer;
|
| 285 |
+
}
|
| 286 |
+
.slider::-moz-range-thumb {
|
| 287 |
+
width: 20px;
|
| 288 |
+
height: 20px;
|
| 289 |
+
border-radius: 50%;
|
| 290 |
+
background: #fff;
|
| 291 |
+
border: 2px solid var(--primary-color);
|
| 292 |
+
cursor: pointer;
|
| 293 |
+
}
|
| 294 |
+
.slider-value {
|
| 295 |
+
font-size: 14px;
|
| 296 |
+
color: var(--text-secondary);
|
| 297 |
+
}
|
| 298 |
+
.textarea-container {
|
| 299 |
+
margin-bottom: 30px;
|
| 300 |
+
}
|
| 301 |
+
.textarea-container label {
|
| 302 |
+
display: block;
|
| 303 |
+
font-size: 18px;
|
| 304 |
+
margin-bottom: 10px;
|
| 305 |
+
}
|
| 306 |
+
.textarea-container textarea {
|
| 307 |
+
width: 100%;
|
| 308 |
+
padding: 10px;
|
| 309 |
+
font-size: 16px;
|
| 310 |
+
border: 2px solid var(--primary-light);
|
| 311 |
+
border-radius: 8px;
|
| 312 |
+
outline: none;
|
| 313 |
+
resize: vertical;
|
| 314 |
+
transition: border-color 0.3s, box-shadow 0.3s;
|
| 315 |
+
box-sizing: border-box;
|
| 316 |
+
height: 200px;
|
| 317 |
+
}
|
| 318 |
+
.textarea-container textarea:focus {
|
| 319 |
+
border-color: var(--primary-color);
|
| 320 |
+
box-shadow: 0 0 0 2px var(--primary-light);
|
| 321 |
+
}
|
| 322 |
+
.voice-group {
|
| 323 |
+
margin-bottom: 20px;
|
| 324 |
+
border: 2px solid var(--primary-light);
|
| 325 |
+
border-radius: 12px;
|
| 326 |
+
overflow: hidden;
|
| 327 |
+
cursor: move;
|
| 328 |
+
background: #fff;
|
| 329 |
+
}
|
| 330 |
+
.voice-header {
|
| 331 |
+
padding: 15px 20px;
|
| 332 |
+
font-size: 18px;
|
| 333 |
+
background: var(--primary-light);
|
| 334 |
+
color: #fff;
|
| 335 |
+
cursor: pointer;
|
| 336 |
+
display: flex;
|
| 337 |
+
justify-content: space-between;
|
| 338 |
+
align-items: center;
|
| 339 |
+
}
|
| 340 |
+
.voice-header:hover {
|
| 341 |
+
background: var(--primary-color);
|
| 342 |
+
}
|
| 343 |
+
.voice-buttons {
|
| 344 |
+
padding: 20px;
|
| 345 |
+
display: none;
|
| 346 |
+
gap: 12px;
|
| 347 |
+
flex-wrap: wrap;
|
| 348 |
+
}
|
| 349 |
+
.voice-button {
|
| 350 |
+
background: var(--secondary-color);
|
| 351 |
+
color: #fff;
|
| 352 |
+
border: none;
|
| 353 |
+
padding: 10px 20px;
|
| 354 |
+
border-radius: 50px;
|
| 355 |
+
cursor: pointer;
|
| 356 |
+
transition: filter 0.3s;
|
| 357 |
+
}
|
| 358 |
+
.voice-button:hover {
|
| 359 |
+
filter: brightness(0.9);
|
| 360 |
+
}
|
| 361 |
+
.chevron {
|
| 362 |
+
transition: transform 0.3s;
|
| 363 |
+
}
|
| 364 |
+
.voice-group.open .voice-buttons {
|
| 365 |
+
display: flex;
|
| 366 |
+
}
|
| 367 |
+
.voice-group.open .chevron {
|
| 368 |
+
transform: rotate(180deg);
|
| 369 |
+
}
|
| 370 |
+
.dragging {
|
| 371 |
+
opacity: 0.5;
|
| 372 |
+
}
|
| 373 |
+
</style>
|
| 374 |
+
</head>
|
| 375 |
+
<body>
|
| 376 |
+
<div class="container">
|
| 377 |
+
<div class="input-area">
|
| 378 |
+
<div class="filter-section">
|
| 379 |
+
<label for="keywords">фильтр по языкам:</label
|
| 380 |
+
><input type="text" id="keywords" value="multilingual,-RU" />
|
| 381 |
+
</div>
|
| 382 |
+
<div class="slider-container">
|
| 383 |
+
<label for="rate">скорость:</label
|
| 384 |
+
><input
|
| 385 |
+
type="range"
|
| 386 |
+
min="-1"
|
| 387 |
+
max="1"
|
| 388 |
+
step="0.1"
|
| 389 |
+
value="-0.1"
|
| 390 |
+
class="slider"
|
| 391 |
+
id="rate"
|
| 392 |
+
/>
|
| 393 |
+
<div class="slider-value" id="rateValue">-0.1</div>
|
| 394 |
+
<label for="pitch">тон:</label
|
| 395 |
+
><input
|
| 396 |
+
type="range"
|
| 397 |
+
min="-1"
|
| 398 |
+
max="1"
|
| 399 |
+
step="0.1"
|
| 400 |
+
value="0.1"
|
| 401 |
+
class="slider"
|
| 402 |
+
id="pitch"
|
| 403 |
+
/>
|
| 404 |
+
<div class="slider-value" id="pitchValue">0.1</div>
|
| 405 |
+
</div>
|
| 406 |
+
<div class="textarea-container">
|
| 407 |
+
<label for="inputText">текст:</label
|
| 408 |
+
><textarea id="inputText">Привет, хочешь я расскажу сказку?</textarea>
|
| 409 |
+
</div>
|
| 410 |
+
</div>
|
| 411 |
+
<div class="output-area">
|
| 412 |
+
<h1>голос</h1>
|
| 413 |
+
<div id="voices"></div>
|
| 414 |
+
</div>
|
| 415 |
+
</div>" +
|
| 416 |
+
"<script>const voiceList = ${JSON.stringify(groupedVoiceList)};let audio=null;function filterVoices(){const keywords=document.getElementById('keywords').value.split(',').map(k=>k.trim().toLowerCase());const voicesDiv=document.getElementById('voices');voicesDiv.innerHTML='';const filteredVoices={};for(const[locale,voices]of Object.entries(voiceList)){const filtered=voices.filter(({name,friendlyName})=>keywords.some(keyword=>name.toLowerCase().includes(keyword)||friendlyName.toLowerCase().includes(keyword)));if(filtered.length>0){filteredVoices[locale]=filtered}}for(const[locale,voices]of Object.entries(filteredVoices)){const group=document.createElement('div');group.className='voice-group';group.draggable=true;const header=document.createElement('div');header.className='voice-header';header.textContent=locale.toUpperCase();const chevron=document.createElement('span');chevron.className='chevron';chevron.innerHTML='▼';header.appendChild(chevron);const buttonsContainer=document.createElement('div');buttonsContainer.className='voice-buttons';voices.forEach(({model,name})=>{const button=document.createElement('button');button.className='voice-button';button.textContent=name;button.onclick=()=>synthesize(model);buttonsContainer.appendChild(button)});header.onclick=()=>{group.classList.toggle('open')};group.appendChild(header);group.appendChild(buttonsContainer);voicesDiv.appendChild(group)}addDragDropListeners()}function synthesize(model){const text=document.getElementById('inputText').value||'Hello world';const rate=document.getElementById('rate').value||'-0.1';const pitch=document.getElementById('pitch').value||'0.1';const voice=\`rate:\${rate}|pitch:\${pitch}\`;if(audio){audio.pause();audio.currentTime=0}fetch('/v1/audio/speech',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({model,input:text,voice})}).then(response=>response.blob()).then(blob=>{const audioUrl=URL.createObjectURL(blob);audio=new Audio(audioUrl);audio.play()})}function addDragDropListeners(){const voicesDiv=document.getElementById('voices');let draggedItem=null;voicesDiv.addEventListener('dragstart',e=>{draggedItem=e.target;e.target.classList.add('dragging')});voicesDiv.addEventListener('dragend',e=>{e.target.classList.remove('dragging');draggedItem=null});voicesDiv.addEventListener('dragover',e=>{e.preventDefault();const afterElement=getDragAfterElement(voicesDiv,e.clientY);if(afterElement==null){voicesDiv.appendChild(draggedItem)}else{voicesDiv.insertBefore(draggedItem,afterElement)}})}function getDragAfterElement(container,y){const draggableElements=[...container.querySelectorAll('.voice-group:not(.dragging)')];return draggableElements.reduce((closest,child)=>{const box=child.getBoundingClientRect();const offset=y-box.top-box.height/2;if(offset<0&&offset>closest.offset){return{offset:offset,element:child}}else{return closest}},{offset:Number.NEGATIVE_INFINITY}).element}filterVoices();document.getElementById('keywords').addEventListener('input',filterVoices);const rateSlider=document.getElementById('rate');const rateValue=document.getElementById('rateValue');rateSlider.oninput=function(){rateValue.innerHTML=this.value};const pitchSlider=document.getElementById('pitch');const pitchValue=document.getElementById('pitchValue');pitchSlider.oninput=function(){pitchValue.innerHTML=this.value}</script>" +
|
| 417 |
+
"</body></html>`;
|
| 418 |
+
|
| 419 |
+
return new Response(html, {
|
| 420 |
+
headers: { "Content-Type": "text/html" },
|
| 421 |
+
});
|
| 422 |
+
}
|
| 423 |
+
|
| 424 |
+
|
| 425 |
+
serve(async (req) => {
|
| 426 |
+
try {
|
| 427 |
+
const url = new URL(req.url);
|
| 428 |
+
|
| 429 |
+
if (url.pathname === "/") {
|
| 430 |
+
return handleDemoRequest(req);
|
| 431 |
+
}
|
| 432 |
+
|
| 433 |
+
if (url.pathname === "/tts") {
|
| 434 |
+
return handleDebugRequest(req);
|
| 435 |
+
}
|
| 436 |
+
|
| 437 |
+
if (url.pathname !== "/v1/audio/speech") {
|
| 438 |
+
console.log(`Unhandled path ${url.pathname}`);
|
| 439 |
+
return new Response("Not Found", { status: 404 });
|
| 440 |
+
}
|
| 441 |
+
|
| 442 |
+
return handleSynthesisRequest(req);
|
| 443 |
+
} catch (err) {
|
| 444 |
+
console.error(`Error processing request: ${err.message}`);
|
| 445 |
+
return new Response(`Internal Server Error\n${err.message}`, {
|
| 446 |
+
status: 500,
|
| 447 |
+
});
|
| 448 |
+
}
|
| 449 |
});
|