Spaces:
Running
Running
Update index.html
Browse files- index.html +108 -46
index.html
CHANGED
@@ -10,15 +10,24 @@
|
|
10 |
<style>
|
11 |
body {
|
12 |
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
max-width: 800px;
|
14 |
margin: 0 auto;
|
15 |
-
padding: 20px;
|
16 |
-
background-color: #f0f0f0;
|
17 |
-
color: #333;
|
18 |
}
|
19 |
h1 {
|
|
|
|
|
|
|
|
|
|
|
20 |
text-align: center;
|
21 |
-
color: #
|
|
|
22 |
}
|
23 |
#chat-container {
|
24 |
display: flex;
|
@@ -27,48 +36,47 @@
|
|
27 |
}
|
28 |
#conversation {
|
29 |
flex-grow: 1;
|
30 |
-
border: 1px solid #
|
31 |
padding: 10px;
|
32 |
overflow-y: scroll;
|
33 |
-
background-color:
|
34 |
border-radius: 5px;
|
35 |
-
|
36 |
}
|
37 |
#controls {
|
38 |
display: flex;
|
39 |
justify-content: center;
|
40 |
-
margin-
|
41 |
}
|
42 |
button {
|
43 |
font-size: 18px;
|
44 |
padding: 10px 20px;
|
45 |
-
|
46 |
-
|
47 |
-
color: white;
|
48 |
border: none;
|
49 |
border-radius: 5px;
|
50 |
cursor: pointer;
|
51 |
transition: background-color 0.3s;
|
52 |
}
|
53 |
button:hover {
|
54 |
-
background-color: #
|
55 |
}
|
56 |
button:disabled {
|
57 |
-
background-color: #
|
58 |
cursor: not-allowed;
|
59 |
}
|
60 |
#visualizer {
|
61 |
width: 100%;
|
62 |
height: 100px;
|
63 |
-
background-color: #
|
64 |
-
margin-top: 20px;
|
65 |
border-radius: 5px;
|
66 |
overflow: hidden;
|
|
|
67 |
}
|
68 |
.bar {
|
69 |
width: 5px;
|
70 |
height: 100%;
|
71 |
-
background-color: #
|
72 |
display: inline-block;
|
73 |
margin-right: 1px;
|
74 |
}
|
@@ -78,7 +86,7 @@
|
|
78 |
left: 0;
|
79 |
width: 100%;
|
80 |
height: 100%;
|
81 |
-
background-color: rgba(0, 0, 0, 0.
|
82 |
display: flex;
|
83 |
justify-content: center;
|
84 |
align-items: center;
|
@@ -88,7 +96,7 @@
|
|
88 |
width: 50px;
|
89 |
height: 50px;
|
90 |
border: 5px solid #f3f3f3;
|
91 |
-
border-top: 5px solid #
|
92 |
border-radius: 50%;
|
93 |
animation: spin 1s linear infinite;
|
94 |
}
|
@@ -96,20 +104,67 @@
|
|
96 |
0% { transform: rotate(0deg); }
|
97 |
100% { transform: rotate(360deg); }
|
98 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
</style>
|
100 |
</head>
|
101 |
<body>
|
102 |
<div id="loading">
|
103 |
<div class="spinner"></div>
|
104 |
</div>
|
105 |
-
<
|
106 |
-
|
107 |
-
<
|
108 |
-
<div id="
|
109 |
-
|
110 |
-
|
111 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
</div>
|
|
|
|
|
|
|
113 |
</div>
|
114 |
|
115 |
<script type="module">
|
@@ -119,9 +174,10 @@
|
|
119 |
|
120 |
const conversationDiv = document.getElementById('conversation');
|
121 |
const startButton = document.getElementById('startButton');
|
122 |
-
const stopButton = document.getElementById('stopButton');
|
123 |
const visualizer = document.getElementById('visualizer');
|
124 |
const loadingDiv = document.getElementById('loading');
|
|
|
|
|
125 |
|
126 |
let myvad;
|
127 |
let sttPipeline;
|
@@ -157,12 +213,12 @@
|
|
157 |
ttsPipeline = await pipeline('text-to-speech', 'Xenova/mms-tts-eng', {
|
158 |
quantized: false,
|
159 |
});
|
160 |
-
|
161 |
startButton.disabled = false;
|
162 |
loadingDiv.style.display = 'none';
|
163 |
} catch (error) {
|
164 |
console.error('Error initializing pipelines:', error);
|
165 |
-
|
166 |
loadingDiv.style.display = 'none';
|
167 |
}
|
168 |
}
|
@@ -174,24 +230,27 @@
|
|
174 |
}
|
175 |
|
176 |
const transcription = await sttPipeline(audio);
|
177 |
-
|
178 |
|
179 |
const botResponse = `I heard you say: "${transcription.text}". This is a placeholder response.`;
|
180 |
-
|
181 |
|
182 |
const speechOutput = await ttsPipeline(botResponse);
|
183 |
playAudio(speechOutput.audio);
|
184 |
} catch (error) {
|
185 |
console.error('Error processing speech:', error);
|
186 |
-
|
187 |
}
|
188 |
}
|
189 |
|
190 |
-
function
|
191 |
-
const
|
192 |
-
|
193 |
-
|
194 |
-
|
|
|
|
|
|
|
195 |
}
|
196 |
|
197 |
function playAudio(audioArray) {
|
@@ -215,11 +274,11 @@
|
|
215 |
|
216 |
myvad = await vad.MicVAD.new({
|
217 |
onSpeechStart: () => {
|
218 |
-
|
219 |
updateVisualizer();
|
220 |
},
|
221 |
onSpeechEnd: (audio) => {
|
222 |
-
|
223 |
cancelAnimationFrame(animationId);
|
224 |
processSpeech(audio);
|
225 |
}
|
@@ -230,27 +289,30 @@
|
|
230 |
source.connect(analyser);
|
231 |
|
232 |
await myvad.start();
|
233 |
-
startButton.
|
234 |
-
|
235 |
-
|
236 |
} catch (error) {
|
237 |
console.error('Error starting VAD:', error);
|
238 |
-
|
239 |
}
|
240 |
}
|
241 |
|
242 |
function stopListening() {
|
243 |
if (myvad) {
|
244 |
myvad.pause();
|
245 |
-
startButton.
|
246 |
-
|
247 |
-
|
248 |
cancelAnimationFrame(animationId);
|
|
|
249 |
}
|
250 |
}
|
251 |
|
252 |
startButton.addEventListener('click', startListening);
|
253 |
-
|
|
|
|
|
254 |
|
255 |
createVisualizer();
|
256 |
initializePipelines();
|
|
|
10 |
<style>
|
11 |
body {
|
12 |
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
|
13 |
+
margin: 0;
|
14 |
+
padding: 20px;
|
15 |
+
background-color: #1a1a1a;
|
16 |
+
color: #f0f0f0;
|
17 |
+
}
|
18 |
+
.container {
|
19 |
max-width: 800px;
|
20 |
margin: 0 auto;
|
|
|
|
|
|
|
21 |
}
|
22 |
h1 {
|
23 |
+
color: #ffd700;
|
24 |
+
text-align: center;
|
25 |
+
margin-bottom: 10px;
|
26 |
+
}
|
27 |
+
.subtitle {
|
28 |
text-align: center;
|
29 |
+
color: #ffd700;
|
30 |
+
margin-bottom: 20px;
|
31 |
}
|
32 |
#chat-container {
|
33 |
display: flex;
|
|
|
36 |
}
|
37 |
#conversation {
|
38 |
flex-grow: 1;
|
39 |
+
border: 1px solid #444;
|
40 |
padding: 10px;
|
41 |
overflow-y: scroll;
|
42 |
+
background-color: #2a2a2a;
|
43 |
border-radius: 5px;
|
44 |
+
margin-bottom: 20px;
|
45 |
}
|
46 |
#controls {
|
47 |
display: flex;
|
48 |
justify-content: center;
|
49 |
+
margin-bottom: 20px;
|
50 |
}
|
51 |
button {
|
52 |
font-size: 18px;
|
53 |
padding: 10px 20px;
|
54 |
+
background-color: #ffd700;
|
55 |
+
color: #1a1a1a;
|
|
|
56 |
border: none;
|
57 |
border-radius: 5px;
|
58 |
cursor: pointer;
|
59 |
transition: background-color 0.3s;
|
60 |
}
|
61 |
button:hover {
|
62 |
+
background-color: #ffec8b;
|
63 |
}
|
64 |
button:disabled {
|
65 |
+
background-color: #666;
|
66 |
cursor: not-allowed;
|
67 |
}
|
68 |
#visualizer {
|
69 |
width: 100%;
|
70 |
height: 100px;
|
71 |
+
background-color: #2a2a2a;
|
|
|
72 |
border-radius: 5px;
|
73 |
overflow: hidden;
|
74 |
+
margin-bottom: 20px;
|
75 |
}
|
76 |
.bar {
|
77 |
width: 5px;
|
78 |
height: 100%;
|
79 |
+
background-color: #ffd700;
|
80 |
display: inline-block;
|
81 |
margin-right: 1px;
|
82 |
}
|
|
|
86 |
left: 0;
|
87 |
width: 100%;
|
88 |
height: 100%;
|
89 |
+
background-color: rgba(0, 0, 0, 0.8);
|
90 |
display: flex;
|
91 |
justify-content: center;
|
92 |
align-items: center;
|
|
|
96 |
width: 50px;
|
97 |
height: 50px;
|
98 |
border: 5px solid #f3f3f3;
|
99 |
+
border-top: 5px solid #ffd700;
|
100 |
border-radius: 50%;
|
101 |
animation: spin 1s linear infinite;
|
102 |
}
|
|
|
104 |
0% { transform: rotate(0deg); }
|
105 |
100% { transform: rotate(360deg); }
|
106 |
}
|
107 |
+
#configuration {
|
108 |
+
margin-bottom: 20px;
|
109 |
+
}
|
110 |
+
select {
|
111 |
+
width: 100%;
|
112 |
+
padding: 10px;
|
113 |
+
font-size: 16px;
|
114 |
+
background-color: #2a2a2a;
|
115 |
+
color: #f0f0f0;
|
116 |
+
border: 1px solid #444;
|
117 |
+
border-radius: 5px;
|
118 |
+
}
|
119 |
+
#model-info {
|
120 |
+
margin-top: 10px;
|
121 |
+
font-size: 14px;
|
122 |
+
color: #aaa;
|
123 |
+
}
|
124 |
+
#logs {
|
125 |
+
background-color: #2a2a2a;
|
126 |
+
border: 1px solid #444;
|
127 |
+
border-radius: 5px;
|
128 |
+
padding: 10px;
|
129 |
+
height: 200px;
|
130 |
+
overflow-y: scroll;
|
131 |
+
font-family: monospace;
|
132 |
+
font-size: 14px;
|
133 |
+
}
|
134 |
+
#clear-logs {
|
135 |
+
margin-top: 10px;
|
136 |
+
font-size: 14px;
|
137 |
+
padding: 5px 10px;
|
138 |
+
}
|
139 |
</style>
|
140 |
</head>
|
141 |
<body>
|
142 |
<div id="loading">
|
143 |
<div class="spinner"></div>
|
144 |
</div>
|
145 |
+
<div class="container">
|
146 |
+
<h1>Voice Chat Bot Demo</h1>
|
147 |
+
<p class="subtitle">For best results, use headphones.</p>
|
148 |
+
<div id="chat-container">
|
149 |
+
<div id="controls">
|
150 |
+
<button id="startButton" disabled>Begin Call</button>
|
151 |
+
</div>
|
152 |
+
<div id="configuration">
|
153 |
+
<select id="configSelect">
|
154 |
+
<option value="fastest">Fastest</option>
|
155 |
+
<option value="balanced">Balanced</option>
|
156 |
+
<option value="quality">Highest Quality</option>
|
157 |
+
</select>
|
158 |
+
<div id="model-info">
|
159 |
+
TTS: Xenova/mms-tts-eng / STT: Xenova/whisper-tiny.en / LLM: Placeholder
|
160 |
+
</div>
|
161 |
+
</div>
|
162 |
+
<div id="visualizer"></div>
|
163 |
+
<div id="conversation"></div>
|
164 |
</div>
|
165 |
+
<h2>Logs</h2>
|
166 |
+
<div id="logs"></div>
|
167 |
+
<button id="clear-logs">Clear</button>
|
168 |
</div>
|
169 |
|
170 |
<script type="module">
|
|
|
174 |
|
175 |
const conversationDiv = document.getElementById('conversation');
|
176 |
const startButton = document.getElementById('startButton');
|
|
|
177 |
const visualizer = document.getElementById('visualizer');
|
178 |
const loadingDiv = document.getElementById('loading');
|
179 |
+
const logsDiv = document.getElementById('logs');
|
180 |
+
const clearLogsButton = document.getElementById('clear-logs');
|
181 |
|
182 |
let myvad;
|
183 |
let sttPipeline;
|
|
|
213 |
ttsPipeline = await pipeline('text-to-speech', 'Xenova/mms-tts-eng', {
|
214 |
quantized: false,
|
215 |
});
|
216 |
+
addLog('System: Voice Chat Bot initialized. Click "Begin Call" to start.');
|
217 |
startButton.disabled = false;
|
218 |
loadingDiv.style.display = 'none';
|
219 |
} catch (error) {
|
220 |
console.error('Error initializing pipelines:', error);
|
221 |
+
addLog('System: Error initializing Voice Chat Bot. Please check the console for details.');
|
222 |
loadingDiv.style.display = 'none';
|
223 |
}
|
224 |
}
|
|
|
230 |
}
|
231 |
|
232 |
const transcription = await sttPipeline(audio);
|
233 |
+
addLog(`User: ${transcription.text}`);
|
234 |
|
235 |
const botResponse = `I heard you say: "${transcription.text}". This is a placeholder response.`;
|
236 |
+
addLog(`Bot: ${botResponse}`);
|
237 |
|
238 |
const speechOutput = await ttsPipeline(botResponse);
|
239 |
playAudio(speechOutput.audio);
|
240 |
} catch (error) {
|
241 |
console.error('Error processing speech:', error);
|
242 |
+
addLog('System: Error processing speech. Please try again.');
|
243 |
}
|
244 |
}
|
245 |
|
246 |
+
function addLog(message) {
|
247 |
+
const now = new Date();
|
248 |
+
const timestamp = now.toLocaleTimeString();
|
249 |
+
const logMessage = `[${timestamp}] ${message}`;
|
250 |
+
const messageElement = document.createElement('div');
|
251 |
+
messageElement.textContent = logMessage;
|
252 |
+
logsDiv.appendChild(messageElement);
|
253 |
+
logsDiv.scrollTop = logsDiv.scrollHeight;
|
254 |
}
|
255 |
|
256 |
function playAudio(audioArray) {
|
|
|
274 |
|
275 |
myvad = await vad.MicVAD.new({
|
276 |
onSpeechStart: () => {
|
277 |
+
addLog('--- vad: speech start');
|
278 |
updateVisualizer();
|
279 |
},
|
280 |
onSpeechEnd: (audio) => {
|
281 |
+
addLog('--- vad: speech end');
|
282 |
cancelAnimationFrame(animationId);
|
283 |
processSpeech(audio);
|
284 |
}
|
|
|
289 |
source.connect(analyser);
|
290 |
|
291 |
await myvad.start();
|
292 |
+
startButton.textContent = 'End Call';
|
293 |
+
startButton.onclick = stopListening;
|
294 |
+
addLog('System: Listening...');
|
295 |
} catch (error) {
|
296 |
console.error('Error starting VAD:', error);
|
297 |
+
addLog('System: Error starting voice detection. Please check your microphone and try again.');
|
298 |
}
|
299 |
}
|
300 |
|
301 |
function stopListening() {
|
302 |
if (myvad) {
|
303 |
myvad.pause();
|
304 |
+
startButton.textContent = 'Begin Call';
|
305 |
+
startButton.onclick = startListening;
|
306 |
+
addLog('System: Stopped listening.');
|
307 |
cancelAnimationFrame(animationId);
|
308 |
+
addLog('websocket closed');
|
309 |
}
|
310 |
}
|
311 |
|
312 |
startButton.addEventListener('click', startListening);
|
313 |
+
clearLogsButton.addEventListener('click', () => {
|
314 |
+
logsDiv.innerHTML = '';
|
315 |
+
});
|
316 |
|
317 |
createVisualizer();
|
318 |
initializePipelines();
|