@creativeorange/azure-text-to-speech 1.0.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/LICENSE +21 -0
- package/README.md +0 -0
- package/dist/co-azure-tts.es.js +6026 -0
- package/dist/co-azure-tts.umd.js +19 -0
- package/index.html +578 -0
- package/package.json +21 -0
- package/src/main.ts +223 -0
- package/tsconfig.json +21 -0
- package/vite.config.ts +20 -0
package/src/main.ts
ADDED
@@ -0,0 +1,223 @@
|
|
1
|
+
import { SpeakerAudioDestination, AudioConfig, OutputFormat, Recognizer, SpeechConfig, SpeechRecognizer, SpeechSynthesizer } from 'microsoft-cognitiveservices-speech-sdk';
|
2
|
+
|
3
|
+
export default class TextToSpeech {
|
4
|
+
|
5
|
+
key: string;
|
6
|
+
region: string;
|
7
|
+
voice: string;
|
8
|
+
|
9
|
+
textToRead: string = '';
|
10
|
+
|
11
|
+
wordBoundryList: any[] = [];
|
12
|
+
|
13
|
+
clickedNode: any;
|
14
|
+
highlightDiv: any;
|
15
|
+
|
16
|
+
speechConfig: any;
|
17
|
+
audioConfig: any;
|
18
|
+
player: any;
|
19
|
+
synthesizer: any;
|
20
|
+
|
21
|
+
previousWordBoundary: any;
|
22
|
+
|
23
|
+
|
24
|
+
constructor(key: string, region: string, voice: string = 'male') {
|
25
|
+
this.key = key;
|
26
|
+
this.region = region;
|
27
|
+
this.voice = voice;
|
28
|
+
}
|
29
|
+
|
30
|
+
async start() {
|
31
|
+
setInterval(() => {
|
32
|
+
if (this.player !== undefined && this.highlightDiv) {
|
33
|
+
const currentTime = this.player.currentTime;
|
34
|
+
let wordBoundary;
|
35
|
+
for (const e of this.wordBoundryList) {
|
36
|
+
if (currentTime * 1000 > e.audioOffset / 10000) {
|
37
|
+
wordBoundary = e;
|
38
|
+
} else {
|
39
|
+
break;
|
40
|
+
}
|
41
|
+
}
|
42
|
+
if (~['.', ',', '!', '?'].indexOf(wordBoundary.text)) {
|
43
|
+
wordBoundary = this.previousWordBoundary ?? undefined;
|
44
|
+
}
|
45
|
+
|
46
|
+
if (wordBoundary !== undefined) {
|
47
|
+
this.previousWordBoundary = wordBoundary;
|
48
|
+
this.highlightDiv.innerHTML = this.textToRead.substr(0, wordBoundary.textOffset) +
|
49
|
+
"<span class='co-tts-highlight'>" + wordBoundary.text + "</span>" +
|
50
|
+
this.textToRead.substr(wordBoundary.textOffset + wordBoundary.wordLength);
|
51
|
+
} else {
|
52
|
+
this.highlightDiv.innerHTML = this.textToRead;
|
53
|
+
}
|
54
|
+
}
|
55
|
+
}, 50);
|
56
|
+
|
57
|
+
await this.registerBindings(document);
|
58
|
+
}
|
59
|
+
|
60
|
+
async synthesis() {
|
61
|
+
|
62
|
+
}
|
63
|
+
|
64
|
+
async registerBindings(node: any) {
|
65
|
+
let nodes = node.childNodes;
|
66
|
+
for (let i = 0; i < nodes.length; i++) {
|
67
|
+
if (!nodes[i]) {
|
68
|
+
continue;
|
69
|
+
}
|
70
|
+
|
71
|
+
let currentNode = nodes[i];
|
72
|
+
|
73
|
+
if (currentNode.attributes) {
|
74
|
+
if (currentNode.attributes.getNamedItem('co-tts.id')) {
|
75
|
+
await this.handleIdModifier(currentNode, currentNode.attributes.getNamedItem('co-tts.id'));
|
76
|
+
} else if (currentNode.attributes.getNamedItem('co-tts.ajax')) {
|
77
|
+
await this.handleAjaxModifier(currentNode, currentNode.attributes.getNamedItem('co-tts.ajax'));
|
78
|
+
} else if (currentNode.attributes.getNamedItem('co-tts')) {
|
79
|
+
await this.handleDefault(currentNode, currentNode.attributes.getNamedItem('co-tts'));
|
80
|
+
} else if (currentNode.attributes.getNamedItem('co-tts.stop')) {
|
81
|
+
await this.handleStopModifier(currentNode, currentNode.attributes.getNamedItem('co-tts.stop'));
|
82
|
+
} else if (currentNode.attributes.getNamedItem('co-tts.resume')) {
|
83
|
+
await this.handleResumeModifier(currentNode, currentNode.attributes.getNamedItem('co-tts.resume'));
|
84
|
+
} else if (currentNode.attributes.getNamedItem('co-tts.pause')) {
|
85
|
+
await this.handlePauseModifier(currentNode, currentNode.attributes.getNamedItem('co-tts.pause'));
|
86
|
+
}
|
87
|
+
}
|
88
|
+
|
89
|
+
if (currentNode.childNodes.length > 0) {
|
90
|
+
await this.registerBindings(currentNode);
|
91
|
+
}
|
92
|
+
}
|
93
|
+
}
|
94
|
+
|
95
|
+
async handleIdModifier(node: any, attr: Attr) {
|
96
|
+
node.addEventListener('click', async (_: any) => {
|
97
|
+
this.stopPlayer();
|
98
|
+
this.clickedNode = node;
|
99
|
+
let referenceDiv = document.getElementById(attr.value);
|
100
|
+
|
101
|
+
if (!referenceDiv) {
|
102
|
+
return;
|
103
|
+
}
|
104
|
+
|
105
|
+
if (referenceDiv.hasAttribute('co-tts.text') && referenceDiv.getAttribute('co-tts.text') !== '') {
|
106
|
+
this.textToRead = referenceDiv.getAttribute('co-tts.text') ?? '';
|
107
|
+
} else {
|
108
|
+
this.textToRead = referenceDiv.innerText;
|
109
|
+
}
|
110
|
+
|
111
|
+
if (referenceDiv.hasAttribute('co-tts.highlight')) {
|
112
|
+
this.highlightDiv = referenceDiv;
|
113
|
+
}
|
114
|
+
|
115
|
+
this.startSynthesizer(node, attr);
|
116
|
+
});
|
117
|
+
}
|
118
|
+
|
119
|
+
async handleAjaxModifier(node: any, attr: Attr) {
|
120
|
+
node.addEventListener('click', async (_: any) => {
|
121
|
+
this.stopPlayer();
|
122
|
+
this.clickedNode = node;
|
123
|
+
let response = await fetch(attr.value, {
|
124
|
+
method: `GET`,
|
125
|
+
});
|
126
|
+
|
127
|
+
this.textToRead = await response.text();
|
128
|
+
|
129
|
+
this.startSynthesizer(node, attr);
|
130
|
+
})
|
131
|
+
}
|
132
|
+
|
133
|
+
async handleDefault(node: any, attr: Attr) {
|
134
|
+
node.addEventListener('click', async (_: any) => {
|
135
|
+
this.stopPlayer();
|
136
|
+
this.clickedNode = node;
|
137
|
+
if (node.hasAttribute('co-tts.highlight')) {
|
138
|
+
this.highlightDiv = node;
|
139
|
+
}
|
140
|
+
if (attr.value === "") {
|
141
|
+
this.textToRead = node.innerText;
|
142
|
+
} else {
|
143
|
+
this.textToRead = attr.value;
|
144
|
+
}
|
145
|
+
|
146
|
+
this.startSynthesizer(node, attr);
|
147
|
+
});
|
148
|
+
}
|
149
|
+
|
150
|
+
async handleStopModifier(node: any, attr: Attr) {
|
151
|
+
node.addEventListener('click', async (_: any) => {
|
152
|
+
await this.stopPlayer();
|
153
|
+
});
|
154
|
+
}
|
155
|
+
|
156
|
+
async handlePauseModifier(node: any, attr: Attr) {
|
157
|
+
node.addEventListener('click', async (_: any) => {
|
158
|
+
await this.player.pause();
|
159
|
+
});
|
160
|
+
}
|
161
|
+
|
162
|
+
async handleResumeModifier(node: any, attr: Attr) {
|
163
|
+
node.addEventListener('click', async (_: any) => {
|
164
|
+
await this.player.resume();
|
165
|
+
});
|
166
|
+
}
|
167
|
+
|
168
|
+
async stopPlayer() {
|
169
|
+
if (this.highlightDiv !== undefined) {
|
170
|
+
this.highlightDiv.innerHTML = this.textToRead;
|
171
|
+
}
|
172
|
+
|
173
|
+
this.textToRead = '';
|
174
|
+
this.wordBoundryList = [];
|
175
|
+
if (this.player !== undefined) {
|
176
|
+
this.player.pause();
|
177
|
+
}
|
178
|
+
this.player = undefined;
|
179
|
+
this.highlightDiv = undefined;
|
180
|
+
}
|
181
|
+
|
182
|
+
async startSynthesizer (node: any, attr: Attr) {
|
183
|
+
this.speechConfig = SpeechConfig.fromSubscription(this.key, this.region);
|
184
|
+
|
185
|
+
if (this.voice === 'female') {
|
186
|
+
this.speechConfig.speechSynthesisVoiceName = 'Microsoft Server Speech Text to Speech Voice (nl-NL, ColetteNeural)';
|
187
|
+
} else {
|
188
|
+
this.speechConfig.speechSynthesisVoiceName = 'Microsoft Server Speech Text to Speech Voice (nl-NL, MaartenNeural)';
|
189
|
+
}
|
190
|
+
this.speechConfig.speechSynthesisOutputFormat = 8;
|
191
|
+
|
192
|
+
this.player = new SpeakerAudioDestination();
|
193
|
+
|
194
|
+
this.audioConfig = AudioConfig.fromSpeakerOutput(this.player);
|
195
|
+
this.synthesizer = new SpeechSynthesizer(this.speechConfig, this.audioConfig);
|
196
|
+
|
197
|
+
this.synthesizer.wordBoundary = (s: any, e: any) => {
|
198
|
+
this.wordBoundryList.push(e);
|
199
|
+
};
|
200
|
+
|
201
|
+
this.player.onAudioEnd = () => {
|
202
|
+
this.stopPlayer();
|
203
|
+
|
204
|
+
if (this.clickedNode.hasAttribute('co-tts.next')) {
|
205
|
+
let nextNode = document.getElementById(this.clickedNode.getAttribute('co-tts.next'));
|
206
|
+
if (nextNode) {
|
207
|
+
nextNode.dispatchEvent(new Event('click'));
|
208
|
+
}
|
209
|
+
}
|
210
|
+
};
|
211
|
+
|
212
|
+
this.synthesizer.speakTextAsync(this.textToRead,
|
213
|
+
() => {
|
214
|
+
this.synthesizer.close();
|
215
|
+
this.synthesizer = undefined;
|
216
|
+
},
|
217
|
+
() => {
|
218
|
+
this.synthesizer.close();
|
219
|
+
this.synthesizer = undefined;
|
220
|
+
});
|
221
|
+
}
|
222
|
+
|
223
|
+
}
|
package/tsconfig.json
ADDED
@@ -0,0 +1,21 @@
|
|
1
|
+
{
|
2
|
+
"compilerOptions": {
|
3
|
+
"target": "ESNext",
|
4
|
+
"useDefineForClassFields": true,
|
5
|
+
"module": "ESNext",
|
6
|
+
"lib": ["ESNext", "DOM"],
|
7
|
+
"moduleResolution": "Node",
|
8
|
+
"strict": true,
|
9
|
+
"sourceMap": true,
|
10
|
+
"resolveJsonModule": true,
|
11
|
+
"isolatedModules": true,
|
12
|
+
"esModuleInterop": true,
|
13
|
+
"noEmit": true,
|
14
|
+
"noUnusedLocals": true,
|
15
|
+
"noUnusedParameters": true,
|
16
|
+
"noImplicitReturns": true,
|
17
|
+
"skipLibCheck": true
|
18
|
+
},
|
19
|
+
"include": ["src"]
|
20
|
+
}
|
21
|
+
|
package/vite.config.ts
ADDED
@@ -0,0 +1,20 @@
|
|
1
|
+
import { defineConfig } from 'vite';
|
2
|
+
import path from 'path';
|
3
|
+
|
4
|
+
// https://vitejs.dev/config/
|
5
|
+
export default defineConfig({
|
6
|
+
build: {
|
7
|
+
lib: {
|
8
|
+
entry: './src/main.ts',
|
9
|
+
name: 'CreativeOrangeAzureTextToSpeech',
|
10
|
+
fileName: 'co-azure-tts',
|
11
|
+
},
|
12
|
+
outDir: 'dist',
|
13
|
+
rollupOptions: {
|
14
|
+
input: "./src/main.ts",
|
15
|
+
output: {
|
16
|
+
dir: "./dist",
|
17
|
+
},
|
18
|
+
}
|
19
|
+
}
|
20
|
+
})
|