@thattobi/narrator-avatar 1.1.10
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/PUBLISHING.md +62 -0
- package/README.md +352 -0
- package/THIRD_PARTY_NOTICES.md +27 -0
- package/dist/NarratorAvatarFull-DhBLV6kE.js +37063 -0
- package/dist/NarratorAvatarFull-DhBLV6kE.js.map +1 -0
- package/dist/index-xRyAk0sW.js +734 -0
- package/dist/index-xRyAk0sW.js.map +1 -0
- package/dist/index.d.ts +72 -0
- package/dist/narrator-avatar.js +6 -0
- package/dist/narrator-avatar.js.map +1 -0
- package/dist/narrator-avatar.umd.cjs +4071 -0
- package/dist/narrator-avatar.umd.cjs.map +1 -0
- package/package.json +59 -0
|
@@ -0,0 +1,734 @@
|
|
|
1
|
+
import { jsx as z } from "react/jsx-runtime";
|
|
2
|
+
import { forwardRef as Gt, useState as Vt, useRef as p, useEffect as et, useCallback as K, useImperativeHandle as qt, lazy as zt, Suspense as Yt } from "react";
|
|
3
|
+
function Tt(t, r) {
|
|
4
|
+
if (t) return !0;
|
|
5
|
+
if (typeof r != "string" || !r.trim()) return !1;
|
|
6
|
+
if (r.startsWith("/") && !r.startsWith("//")) return !0;
|
|
7
|
+
if (typeof window > "u") return !1;
|
|
8
|
+
try {
|
|
9
|
+
const o = new URL(r, window.location.href);
|
|
10
|
+
return o.origin === window.location.origin && /^https?:$/.test(o.protocol);
|
|
11
|
+
} catch {
|
|
12
|
+
return !1;
|
|
13
|
+
}
|
|
14
|
+
}
|
|
15
|
+
const S = {}, Ct = "https://api.deepgram.com/v1/speak", Jt = "encoding=linear16&container=wav&sample_rate=24000", At = 44, Ht = 3, Zt = 280, mt = 560;
|
|
16
|
+
function kt(t) {
|
|
17
|
+
return typeof t == "string" ? t.trim() : "";
|
|
18
|
+
}
|
|
19
|
+
function Pt(t) {
|
|
20
|
+
return t === "deepgram";
|
|
21
|
+
}
|
|
22
|
+
const rt = (t) => new Promise((r) => setTimeout(r, t));
|
|
23
|
+
function Qt(t) {
|
|
24
|
+
const r = t.trim();
|
|
25
|
+
return r ? r.split(new RegExp("(?<=[.!?])\\s+")).map((o) => o.trim()).filter(Boolean) : [];
|
|
26
|
+
}
|
|
27
|
+
function Xt(t, {
|
|
28
|
+
tinyMaxWords: r = 4,
|
|
29
|
+
tinyMaxChars: o = 48,
|
|
30
|
+
maxChars: e = 280
|
|
31
|
+
} = {}) {
|
|
32
|
+
if (!(t != null && t.length)) return [];
|
|
33
|
+
const n = [];
|
|
34
|
+
let i = "";
|
|
35
|
+
const s = (c) => c.trim().split(/\s+/).filter(Boolean).length, a = (c) => {
|
|
36
|
+
const l = (c || "").trim();
|
|
37
|
+
return !!l && s(l) <= r && l.length <= o;
|
|
38
|
+
}, y = (c) => /[.!?…]["')\]]*\s*$/.test((c || "").trim());
|
|
39
|
+
for (const c of t) {
|
|
40
|
+
if (!i) {
|
|
41
|
+
i = c;
|
|
42
|
+
continue;
|
|
43
|
+
}
|
|
44
|
+
const l = `${i} ${c}`, f = (a(i) || a(c)) && l.length <= e, m = !y(i) && l.length <= e;
|
|
45
|
+
f || m ? i = l : (n.push(i), i = c);
|
|
46
|
+
}
|
|
47
|
+
return i && n.push(i), n;
|
|
48
|
+
}
|
|
49
|
+
function Et(t) {
|
|
50
|
+
return Xt(Qt(t));
|
|
51
|
+
}
|
|
52
|
+
function Rt(t, r = mt) {
|
|
53
|
+
const o = (t || "").trim();
|
|
54
|
+
let e = Math.max(320, Number(r) || mt);
|
|
55
|
+
/\?\s*$/.test(o) ? e += 220 : /!\s*$/.test(o) ? e += 160 : /(?:\.\.\.|…)\s*$/.test(o) ? e += 240 : /\.\s*$/.test(o) && (e += 120);
|
|
56
|
+
const n = o.split(/\s+/).filter(Boolean).length;
|
|
57
|
+
return n >= 24 ? e += 120 : n >= 14 && (e += 60), e += Math.floor(Math.random() * 100) - 30, Math.max(380, Math.min(1100, Math.round(e)));
|
|
58
|
+
}
|
|
59
|
+
async function te(t, r, o) {
|
|
60
|
+
const e = Date.now() + Math.max(0, t);
|
|
61
|
+
for (; Date.now() < e; ) {
|
|
62
|
+
if (!r()) return;
|
|
63
|
+
for (; o != null && o.current && r(); ) await rt(80);
|
|
64
|
+
if (!r()) return;
|
|
65
|
+
await rt(Math.min(60, e - Date.now()));
|
|
66
|
+
}
|
|
67
|
+
}
|
|
68
|
+
function ee(t) {
|
|
69
|
+
return t == null ? "" : Array.isArray(t) ? t.join("").replace(/\s+/g, " ").trim() : typeof t == "string" ? t.replace(/\s+/g, " ").trim() : String(t).replace(/\s+/g, " ").trim();
|
|
70
|
+
}
|
|
71
|
+
function re(t, r = "", o = 0) {
|
|
72
|
+
let e = (r || "").trim();
|
|
73
|
+
const n = () => {
|
|
74
|
+
if (!t) return;
|
|
75
|
+
const i = Number(o) || 0;
|
|
76
|
+
if (i <= 0) {
|
|
77
|
+
t(e);
|
|
78
|
+
return;
|
|
79
|
+
}
|
|
80
|
+
const s = e.split(/\s+/).filter(Boolean);
|
|
81
|
+
t(s.length <= i ? e : s.slice(-i).join(" "));
|
|
82
|
+
};
|
|
83
|
+
return e && n(), (i) => {
|
|
84
|
+
const s = ee(i);
|
|
85
|
+
s && (e ? s.startsWith(e) && s.length > e.length ? e = s : e === s || e.endsWith(` ${s}`) || e.endsWith(s) || (e = `${e} ${s}`) : e = s, n());
|
|
86
|
+
};
|
|
87
|
+
}
|
|
88
|
+
function ne(t, r) {
|
|
89
|
+
var e;
|
|
90
|
+
if (!((e = t == null ? void 0 : t.words) != null && e.length) || r <= 0) return "";
|
|
91
|
+
const o = [];
|
|
92
|
+
for (let n = 0; n < t.words.length; n++) {
|
|
93
|
+
const i = t.wtimes[n];
|
|
94
|
+
if (i + t.wdurations[n] <= r + 1 || i < r) o.push(t.words[n]);
|
|
95
|
+
else break;
|
|
96
|
+
}
|
|
97
|
+
return o.join(" ");
|
|
98
|
+
}
|
|
99
|
+
function oe(t, r) {
|
|
100
|
+
if (r.byteLength < At) return null;
|
|
101
|
+
const o = new DataView(r), e = o.getUint32(24, !0), n = Math.max(1, o.getUint16(22, !0)), i = r.byteLength - At, s = i / (2 * n), a = t.createBuffer(n, s, e), y = new Int16Array(r, At, i / 2 | 0);
|
|
102
|
+
for (let c = 0; c < n; c++) {
|
|
103
|
+
const l = a.getChannelData(c);
|
|
104
|
+
for (let f = 0; f < s; f++)
|
|
105
|
+
l[f] = y[f * n + c] / 32768;
|
|
106
|
+
}
|
|
107
|
+
return a;
|
|
108
|
+
}
|
|
109
|
+
function ie(t) {
|
|
110
|
+
const r = t.replace(/[^\p{L}\p{N}'’-]/gu, ""), o = r.replace(/[^\p{L}]/gu, ""), e = (o.match(/[aeiouy]/gi) || []).length;
|
|
111
|
+
return Math.max(1, o.length + e * 0.65 + (r.length - o.length) * 0.45);
|
|
112
|
+
}
|
|
113
|
+
function se(t) {
|
|
114
|
+
return /[.?!…]["')\]}]*$/.test(t) ? 175 : /[;:]["')\]}]*$/.test(t) ? 115 : /[,—-]["')\]}]*$/.test(t) ? 75 : 0;
|
|
115
|
+
}
|
|
116
|
+
function ue(t, r) {
|
|
117
|
+
const o = t.trim().split(/\s+/).filter(Boolean), e = Math.max(0, Number(r) || 0), n = o.map(se), i = n.reduce((M, C) => M + C, 0), s = i > 0 ? Math.min(1, e * 0.28 / i) : 0, a = i * s, y = Math.max(0, e - a), c = o.map(ie), l = c.reduce((M, C) => M + C, 0) || 1, f = [], m = [];
|
|
118
|
+
let E = 0;
|
|
119
|
+
for (let M = 0; M < o.length; M++) {
|
|
120
|
+
f.push(E);
|
|
121
|
+
const C = c[M] / l * y;
|
|
122
|
+
m.push(C), E += C + n[M] * s;
|
|
123
|
+
}
|
|
124
|
+
return f.length || (o.push(t || " "), f.push(0), m.push(e)), { words: o, wtimes: f, wdurations: m };
|
|
125
|
+
}
|
|
126
|
+
async function ae(t, r, o, e, n = Ct, i = 1) {
|
|
127
|
+
const s = n.includes("?") ? "&" : "?", a = Math.max(0.7, Math.min(1.5, Number(i) || 1)), y = `${n}${s}model=${encodeURIComponent(r)}&speed=${a}&${Jt}`;
|
|
128
|
+
let c;
|
|
129
|
+
for (let l = 0; l < Ht; l++)
|
|
130
|
+
try {
|
|
131
|
+
const f = { "Content-Type": "application/json", Accept: "audio/wav" };
|
|
132
|
+
t && (f.Authorization = `Token ${t}`);
|
|
133
|
+
const m = await fetch(y, {
|
|
134
|
+
method: "POST",
|
|
135
|
+
headers: f,
|
|
136
|
+
body: JSON.stringify({ text: o }),
|
|
137
|
+
signal: e
|
|
138
|
+
});
|
|
139
|
+
if (!m.ok) {
|
|
140
|
+
const M = (await m.text()).slice(0, 500);
|
|
141
|
+
throw new Error(
|
|
142
|
+
`Deepgram TTS error: ${m.status} ${m.statusText}${M ? ` — ${M}` : ""}`
|
|
143
|
+
);
|
|
144
|
+
}
|
|
145
|
+
const E = await m.arrayBuffer();
|
|
146
|
+
if (!(E != null && E.byteLength)) throw new Error("Deepgram TTS returned empty audio");
|
|
147
|
+
return E;
|
|
148
|
+
} catch (f) {
|
|
149
|
+
if (c = f, (f == null ? void 0 : f.name) === "AbortError" || e != null && e.aborted) throw f;
|
|
150
|
+
l < Ht - 1 && await rt(Zt * (l + 1));
|
|
151
|
+
}
|
|
152
|
+
throw c instanceof TypeError && /failed to fetch/i.test(c.message) ? new Error(
|
|
153
|
+
"Deepgram TTS request could not reach the API. Check the configured same-origin proxy/service and browser network policy."
|
|
154
|
+
) : c || new Error("Deepgram TTS failed");
|
|
155
|
+
}
|
|
156
|
+
async function ce(t, r, o, e, n, i = 1, s) {
|
|
157
|
+
const a = await ae(r, o, e, n, s, i);
|
|
158
|
+
let y = oe(t, a);
|
|
159
|
+
if (!y) throw new Error("Failed to prepare audio");
|
|
160
|
+
const c = y.duration * 1e3, { words: l, wtimes: f, wdurations: m } = ue(e, c);
|
|
161
|
+
return { phrase: e, audioBuffer: y, durationMs: c, words: l, wtimes: f, wdurations: m };
|
|
162
|
+
}
|
|
163
|
+
function Wt(t, r, o, e = "") {
|
|
164
|
+
const n = Math.max(0.7, Math.min(1.5, Number(r) || 1));
|
|
165
|
+
return `deepgram|${e}|${t}|${n}|${o}`;
|
|
166
|
+
}
|
|
167
|
+
function ht(t, r, o, e, n, i, s = 1, a) {
|
|
168
|
+
const y = Wt(e, s, n, a), c = t.get(y);
|
|
169
|
+
if (c)
|
|
170
|
+
return Promise.resolve(c).then(
|
|
171
|
+
(f) => f,
|
|
172
|
+
() => (t.delete(y), ht(t, r, o, e, n, null, s, a))
|
|
173
|
+
);
|
|
174
|
+
const l = ce(r, o, e, n, null, s, a).catch((f) => {
|
|
175
|
+
throw t.delete(y), f;
|
|
176
|
+
});
|
|
177
|
+
return t.set(y, l), l;
|
|
178
|
+
}
|
|
179
|
+
function Ft(t, r) {
|
|
180
|
+
if (!t || t.size <= r) return;
|
|
181
|
+
const o = [...t.keys()];
|
|
182
|
+
for (const e of o.slice(0, t.size - r)) t.delete(e);
|
|
183
|
+
}
|
|
184
|
+
function le(t, r, o) {
|
|
185
|
+
if (!r || !t) return null;
|
|
186
|
+
const e = r.sampleRate;
|
|
187
|
+
let n = Math.floor(Math.max(0, o) / 1e3 * e);
|
|
188
|
+
if (n < 0 && (n = 0), n >= r.length - 32) return null;
|
|
189
|
+
const i = r.length - n, s = t.createBuffer(r.numberOfChannels, i, e);
|
|
190
|
+
for (let a = 0; a < r.numberOfChannels; a++)
|
|
191
|
+
s.getChannelData(a).set(r.getChannelData(a).subarray(n));
|
|
192
|
+
return s;
|
|
193
|
+
}
|
|
194
|
+
function fe(t, r, o, e) {
|
|
195
|
+
const n = [], i = [], s = [];
|
|
196
|
+
for (let a = 0; a < t.length; a++) {
|
|
197
|
+
const y = r[a], c = o[a], l = y + c;
|
|
198
|
+
l <= e + 1 || (y < e ? (n.push(t[a]), i.push(0), s.push(l - e)) : (n.push(t[a]), i.push(y - e), s.push(c)));
|
|
199
|
+
}
|
|
200
|
+
return n.length || (n.push(" "), i.push(0), s.push(120)), { words: n, wtimes: i, wdurations: s };
|
|
201
|
+
}
|
|
202
|
+
function de(t, r, o) {
|
|
203
|
+
const e = le(t, r.audioBuffer, o);
|
|
204
|
+
if (!e) return null;
|
|
205
|
+
const { words: n, wtimes: i, wdurations: s } = fe(
|
|
206
|
+
r.words,
|
|
207
|
+
r.wtimes,
|
|
208
|
+
r.wdurations,
|
|
209
|
+
o
|
|
210
|
+
);
|
|
211
|
+
return { audio: e, words: n, wtimes: i, wdurations: s, durationMs: e.duration * 1e3 };
|
|
212
|
+
}
|
|
213
|
+
async function Y(t) {
|
|
214
|
+
if (!t || t.state === "closed") return !1;
|
|
215
|
+
if (t.state === "running") return !0;
|
|
216
|
+
try {
|
|
217
|
+
await Promise.race([
|
|
218
|
+
t.resume(),
|
|
219
|
+
new Promise((r, o) => setTimeout(() => o(new Error("timeout")), 1500))
|
|
220
|
+
]);
|
|
221
|
+
} catch {
|
|
222
|
+
try {
|
|
223
|
+
await t.resume();
|
|
224
|
+
} catch {
|
|
225
|
+
return !1;
|
|
226
|
+
}
|
|
227
|
+
}
|
|
228
|
+
return t.state === "running";
|
|
229
|
+
}
|
|
230
|
+
function he(t, r, {
|
|
231
|
+
onWord: o,
|
|
232
|
+
abortedRef: e,
|
|
233
|
+
pausedRef: n,
|
|
234
|
+
sessionRef: i,
|
|
235
|
+
sessionId: s,
|
|
236
|
+
playbackState: a
|
|
237
|
+
}) {
|
|
238
|
+
return new Promise((y) => {
|
|
239
|
+
const c = t.audioCtx;
|
|
240
|
+
let l = null, f = !1;
|
|
241
|
+
const m = (E) => {
|
|
242
|
+
if (!f) {
|
|
243
|
+
f = !0, l != null && (clearInterval(l), l = null);
|
|
244
|
+
for (const M of t.wordTimers || []) clearTimeout(M);
|
|
245
|
+
if (t.wordTimers = [], t.isAudioPlaying = !1, t.source)
|
|
246
|
+
try {
|
|
247
|
+
t.source.onended = null;
|
|
248
|
+
} catch {
|
|
249
|
+
}
|
|
250
|
+
E !== "paused" && (t.isSpeaking = !1), y(E);
|
|
251
|
+
}
|
|
252
|
+
};
|
|
253
|
+
(async () => {
|
|
254
|
+
var C, nt, H;
|
|
255
|
+
if (!c || !(r != null && r.audio)) {
|
|
256
|
+
m("aborted");
|
|
257
|
+
return;
|
|
258
|
+
}
|
|
259
|
+
if (!await Y(c)) {
|
|
260
|
+
m("blocked");
|
|
261
|
+
return;
|
|
262
|
+
}
|
|
263
|
+
if (e.current || i.current !== s) {
|
|
264
|
+
m("aborted");
|
|
265
|
+
return;
|
|
266
|
+
}
|
|
267
|
+
if (n.current) {
|
|
268
|
+
m("paused");
|
|
269
|
+
return;
|
|
270
|
+
}
|
|
271
|
+
try {
|
|
272
|
+
(nt = (C = t.source) == null ? void 0 : C.stop) == null || nt.call(C);
|
|
273
|
+
} catch {
|
|
274
|
+
}
|
|
275
|
+
t.source = null;
|
|
276
|
+
for (const b of t.wordTimers || []) clearTimeout(b);
|
|
277
|
+
t.wordTimers = [];
|
|
278
|
+
const M = c.createBufferSource();
|
|
279
|
+
if (M.buffer = r.audio, M.connect(t.gain), t.source = M, t.isSpeaking = !0, t.isAudioPlaying = !0, a.segmentStartedCtxTime = c.currentTime, typeof o == "function" && ((H = r.words) != null && H.length))
|
|
280
|
+
for (let b = 0; b < r.words.length; b++) {
|
|
281
|
+
const U = Math.max(0, r.wtimes[b]), ot = r.words[b], P = setTimeout(() => {
|
|
282
|
+
e.current || n.current || i.current !== s || o(ot);
|
|
283
|
+
}, U);
|
|
284
|
+
t.wordTimers.push(P);
|
|
285
|
+
}
|
|
286
|
+
M.onended = () => {
|
|
287
|
+
if (n.current) {
|
|
288
|
+
m("paused");
|
|
289
|
+
return;
|
|
290
|
+
}
|
|
291
|
+
e.current || i.current !== s ? m("aborted") : m("done");
|
|
292
|
+
}, l = setInterval(() => {
|
|
293
|
+
if (!f) {
|
|
294
|
+
if (e.current || i.current !== s) {
|
|
295
|
+
try {
|
|
296
|
+
M.stop();
|
|
297
|
+
} catch {
|
|
298
|
+
}
|
|
299
|
+
m("aborted");
|
|
300
|
+
return;
|
|
301
|
+
}
|
|
302
|
+
if (n.current) {
|
|
303
|
+
try {
|
|
304
|
+
const b = (c.currentTime - a.segmentStartedCtxTime) * 1e3;
|
|
305
|
+
a.resumeOffsetMs = (a.segmentOffsetMs || 0) + Math.max(0, b), M.stop();
|
|
306
|
+
} catch {
|
|
307
|
+
}
|
|
308
|
+
m("paused");
|
|
309
|
+
}
|
|
310
|
+
}
|
|
311
|
+
}, 40);
|
|
312
|
+
try {
|
|
313
|
+
M.start(0);
|
|
314
|
+
} catch {
|
|
315
|
+
m("aborted");
|
|
316
|
+
}
|
|
317
|
+
})();
|
|
318
|
+
});
|
|
319
|
+
}
|
|
320
|
+
const _t = Gt(function({
|
|
321
|
+
ttsApiKey: r = null,
|
|
322
|
+
deepgramEndpoint: o = Ct,
|
|
323
|
+
ttsVoice: e = "aura-2-thalia-en",
|
|
324
|
+
ttsService: n = "deepgram",
|
|
325
|
+
speechRate: i = 1,
|
|
326
|
+
phraseGapMs: s = mt,
|
|
327
|
+
preloadText: a = null,
|
|
328
|
+
onPreloadReady: y = null,
|
|
329
|
+
pauseWhenHidden: c = !0,
|
|
330
|
+
subtitleMaxWords: l = 16,
|
|
331
|
+
onReady: f = () => {
|
|
332
|
+
},
|
|
333
|
+
onError: m = () => {
|
|
334
|
+
},
|
|
335
|
+
onSpeechStart: E = () => {
|
|
336
|
+
},
|
|
337
|
+
onSpeechEnd: M = () => {
|
|
338
|
+
},
|
|
339
|
+
onSubtitle: C = null,
|
|
340
|
+
className: nt = "",
|
|
341
|
+
style: H = {}
|
|
342
|
+
}, b) {
|
|
343
|
+
const [U, ot] = Vt(!1), P = p(null), F = p(/* @__PURE__ */ new Map()), J = p(0), Z = p(!1), O = p(!1), j = p(!1), G = p({
|
|
344
|
+
segmentOffsetMs: 0,
|
|
345
|
+
segmentStartedCtxTime: null,
|
|
346
|
+
durationMs: 0,
|
|
347
|
+
resumeOffsetMs: 0
|
|
348
|
+
}), wt = p(0), Q = p(!1), V = p(r), I = p(o), it = p(e), st = p(i), bt = p(s), Kt = p(a), ut = p(y), W = p(f), D = p(m), at = p(E), L = p(M), $ = p(C), xt = p(l), Nt = p(c), ct = p(() => {
|
|
349
|
+
}), lt = p(async () => {
|
|
350
|
+
}), ft = p(() => {
|
|
351
|
+
});
|
|
352
|
+
et(() => {
|
|
353
|
+
V.current = r, I.current = o || Ct, it.current = e, st.current = Math.max(0.6, Math.min(1.2, Number(i) || 1));
|
|
354
|
+
const u = Number(s);
|
|
355
|
+
bt.current = Number.isFinite(u) && u >= 0 ? u : mt, Kt.current = a, ut.current = y, W.current = f, D.current = m, at.current = E, L.current = M, $.current = C, xt.current = Number(l) > 0 ? Number(l) : 16, Nt.current = c;
|
|
356
|
+
}), et(() => {
|
|
357
|
+
var T, A;
|
|
358
|
+
if (!Pt(n)) {
|
|
359
|
+
ot(!0), (T = W.current) == null || T.call(W);
|
|
360
|
+
return;
|
|
361
|
+
}
|
|
362
|
+
const u = window.AudioContext || window.webkitAudioContext, d = new u(), w = d.createGain();
|
|
363
|
+
return w.connect(d.destination), P.current = {
|
|
364
|
+
audioCtx: d,
|
|
365
|
+
gain: w,
|
|
366
|
+
source: null,
|
|
367
|
+
wordTimers: [],
|
|
368
|
+
isSpeaking: !1,
|
|
369
|
+
isAudioPlaying: !1
|
|
370
|
+
}, ot(!0), (A = W.current) == null || A.call(W), () => {
|
|
371
|
+
var k, g, v, h;
|
|
372
|
+
(k = ft.current) == null || k.call(ft, !0);
|
|
373
|
+
try {
|
|
374
|
+
(h = (v = (g = P.current) == null ? void 0 : g.source) == null ? void 0 : v.stop) == null || h.call(v);
|
|
375
|
+
} catch {
|
|
376
|
+
}
|
|
377
|
+
try {
|
|
378
|
+
d.close();
|
|
379
|
+
} catch {
|
|
380
|
+
}
|
|
381
|
+
P.current = null, F.current.clear();
|
|
382
|
+
};
|
|
383
|
+
}, [n]), et(() => {
|
|
384
|
+
const u = () => {
|
|
385
|
+
var v, h;
|
|
386
|
+
if (!Nt.current) return;
|
|
387
|
+
const g = P.current;
|
|
388
|
+
g && (g.isSpeaking || g.isAudioPlaying) && !O.current && (j.current = !0, (v = ct.current) == null || v.call(ct));
|
|
389
|
+
try {
|
|
390
|
+
((h = g == null ? void 0 : g.audioCtx) == null ? void 0 : h.state) === "running" && g.audioCtx.suspend();
|
|
391
|
+
} catch {
|
|
392
|
+
}
|
|
393
|
+
}, d = async () => {
|
|
394
|
+
var h;
|
|
395
|
+
const g = P.current;
|
|
396
|
+
!g || !await Y(g.audioCtx) || j.current && (j.current = !1, await ((h = lt.current) == null ? void 0 : h.call(lt)));
|
|
397
|
+
}, w = () => {
|
|
398
|
+
document.hidden ? u() : d();
|
|
399
|
+
}, T = () => u(), A = () => {
|
|
400
|
+
var g;
|
|
401
|
+
j.current ? d() : Y((g = P.current) == null ? void 0 : g.audioCtx);
|
|
402
|
+
}, k = () => {
|
|
403
|
+
d();
|
|
404
|
+
};
|
|
405
|
+
return document.addEventListener("visibilitychange", w), window.addEventListener("pagehide", T), window.addEventListener("pageshow", k), window.addEventListener("pointerdown", A, { passive: !0 }), window.addEventListener("touchstart", A, { passive: !0 }), () => {
|
|
406
|
+
document.removeEventListener("visibilitychange", w), window.removeEventListener("pagehide", T), window.removeEventListener("pageshow", k), window.removeEventListener("pointerdown", A), window.removeEventListener("touchstart", A);
|
|
407
|
+
};
|
|
408
|
+
}, []);
|
|
409
|
+
const X = K((u = !0) => {
|
|
410
|
+
var w, T, A;
|
|
411
|
+
Z.current = !0, O.current = !1, Q.current = !1, u && (J.current += 1);
|
|
412
|
+
const d = P.current;
|
|
413
|
+
if (d) {
|
|
414
|
+
for (const k of d.wordTimers || []) clearTimeout(k);
|
|
415
|
+
d.wordTimers = [];
|
|
416
|
+
try {
|
|
417
|
+
(T = (w = d.source) == null ? void 0 : w.stop) == null || T.call(w);
|
|
418
|
+
} catch {
|
|
419
|
+
}
|
|
420
|
+
d.source = null, d.isSpeaking = !1, d.isAudioPlaying = !1;
|
|
421
|
+
}
|
|
422
|
+
try {
|
|
423
|
+
(A = $.current) == null || A.call($, "");
|
|
424
|
+
} catch {
|
|
425
|
+
}
|
|
426
|
+
}, []);
|
|
427
|
+
ft.current = X;
|
|
428
|
+
const dt = K((u) => {
|
|
429
|
+
const d = P.current;
|
|
430
|
+
if (!(d != null && d.audioCtx) || !(u != null && u.trim()) || !Pt(n)) return;
|
|
431
|
+
const w = kt(
|
|
432
|
+
n === "deepgram" ? V.current || (S == null ? void 0 : S.VITE_DEEPGRAM_API_KEY) : V.current
|
|
433
|
+
);
|
|
434
|
+
if (n === "deepgram" && !Tt(w, I.current)) return;
|
|
435
|
+
const T = it.current || "aura-2-thalia-en", A = st.current, k = Et(u), g = k.length ? k : [u.trim()], v = ++wt.current, h = (x) => ht(
|
|
436
|
+
F.current,
|
|
437
|
+
d.audioCtx,
|
|
438
|
+
w,
|
|
439
|
+
T,
|
|
440
|
+
g[x],
|
|
441
|
+
null,
|
|
442
|
+
A,
|
|
443
|
+
I.current
|
|
444
|
+
);
|
|
445
|
+
(async () => {
|
|
446
|
+
var x;
|
|
447
|
+
try {
|
|
448
|
+
if (await h(0), v !== wt.current) return;
|
|
449
|
+
(x = ut.current) == null || x.call(ut);
|
|
450
|
+
} catch (q) {
|
|
451
|
+
console.warn("[VoiceOnlyNarrator] Preload failed:", q);
|
|
452
|
+
return;
|
|
453
|
+
}
|
|
454
|
+
v === wt.current && (g[1] && h(1), Ft(F.current, 8));
|
|
455
|
+
})();
|
|
456
|
+
}, [n]);
|
|
457
|
+
et(() => {
|
|
458
|
+
!U || !a || dt(a);
|
|
459
|
+
}, [U, a, dt]);
|
|
460
|
+
const $t = K(async (u, d) => {
|
|
461
|
+
var Lt, St, It;
|
|
462
|
+
const w = P.current;
|
|
463
|
+
if (!(w != null && w.audioCtx)) return;
|
|
464
|
+
const T = kt(
|
|
465
|
+
n === "deepgram" ? V.current || (S == null ? void 0 : S.VITE_DEEPGRAM_API_KEY) : V.current
|
|
466
|
+
);
|
|
467
|
+
if (n === "deepgram" && !Tt(T, I.current))
|
|
468
|
+
throw new Error("Deepgram requires an API key or a same-origin proxy endpoint");
|
|
469
|
+
const A = it.current || "aura-2-thalia-en", k = st.current, g = Et(u), v = g.length ? g : [u.trim() || " "], h = () => !Z.current && J.current === d, x = /* @__PURE__ */ new Map(), q = (N) => {
|
|
470
|
+
N < 0 || N >= v.length || x.has(N) || x.set(
|
|
471
|
+
N,
|
|
472
|
+
ht(
|
|
473
|
+
F.current,
|
|
474
|
+
w.audioCtx,
|
|
475
|
+
T,
|
|
476
|
+
A,
|
|
477
|
+
v[N],
|
|
478
|
+
null,
|
|
479
|
+
k,
|
|
480
|
+
I.current
|
|
481
|
+
)
|
|
482
|
+
);
|
|
483
|
+
};
|
|
484
|
+
if (q(0), !await Y(w.audioCtx))
|
|
485
|
+
throw new Error("AudioContext blocked — tap Speak / Continue Learning to unlock audio");
|
|
486
|
+
if (await x.get(0), !h()) return;
|
|
487
|
+
q(1);
|
|
488
|
+
let yt = !1, Mt = !1;
|
|
489
|
+
for (let N = 0; N < v.length && h(); N++) {
|
|
490
|
+
for (; O.current && h(); ) await rt(80);
|
|
491
|
+
if (!h()) break;
|
|
492
|
+
q(N), q(N + 1);
|
|
493
|
+
let B;
|
|
494
|
+
try {
|
|
495
|
+
B = await x.get(N);
|
|
496
|
+
} catch (tt) {
|
|
497
|
+
(Lt = D.current) == null || Lt.call(D, tt), x.delete(N);
|
|
498
|
+
continue;
|
|
499
|
+
}
|
|
500
|
+
if (x.delete(N), !B || !h()) break;
|
|
501
|
+
(St = $.current) == null || St.call($, "");
|
|
502
|
+
let _ = 0, R = !1;
|
|
503
|
+
for (; !R && h(); ) {
|
|
504
|
+
for (; O.current && h(); ) await rt(80);
|
|
505
|
+
if (!h()) break;
|
|
506
|
+
if (_ >= B.durationMs - 60) {
|
|
507
|
+
R = !0;
|
|
508
|
+
break;
|
|
509
|
+
}
|
|
510
|
+
const tt = de(w.audioCtx, B, _);
|
|
511
|
+
if (!tt) {
|
|
512
|
+
R = !0;
|
|
513
|
+
break;
|
|
514
|
+
}
|
|
515
|
+
G.current = {
|
|
516
|
+
segmentOffsetMs: _,
|
|
517
|
+
segmentStartedCtxTime: null,
|
|
518
|
+
durationMs: B.durationMs,
|
|
519
|
+
resumeOffsetMs: _
|
|
520
|
+
};
|
|
521
|
+
const Ut = re(
|
|
522
|
+
(jt) => {
|
|
523
|
+
var Bt;
|
|
524
|
+
return (Bt = $.current) == null ? void 0 : Bt.call($, jt);
|
|
525
|
+
},
|
|
526
|
+
ne(B, _),
|
|
527
|
+
xt.current
|
|
528
|
+
);
|
|
529
|
+
Q.current = !0;
|
|
530
|
+
const vt = await he(w, tt, {
|
|
531
|
+
onWord: Ut,
|
|
532
|
+
abortedRef: Z,
|
|
533
|
+
pausedRef: O,
|
|
534
|
+
sessionRef: J,
|
|
535
|
+
sessionId: d,
|
|
536
|
+
playbackState: G.current
|
|
537
|
+
});
|
|
538
|
+
if (vt === "blocked") {
|
|
539
|
+
Mt = !0;
|
|
540
|
+
break;
|
|
541
|
+
}
|
|
542
|
+
if (vt === "paused") {
|
|
543
|
+
_ = G.current.resumeOffsetMs || _;
|
|
544
|
+
continue;
|
|
545
|
+
}
|
|
546
|
+
if (vt === "aborted") break;
|
|
547
|
+
yt = !0, R = !0;
|
|
548
|
+
}
|
|
549
|
+
if (Mt) break;
|
|
550
|
+
try {
|
|
551
|
+
F.current.delete(Wt(
|
|
552
|
+
A,
|
|
553
|
+
k,
|
|
554
|
+
B.phrase,
|
|
555
|
+
I.current
|
|
556
|
+
));
|
|
557
|
+
} catch {
|
|
558
|
+
}
|
|
559
|
+
Ft(F.current, 8), h() && R && N < v.length - 1 && await te(
|
|
560
|
+
Rt(B.phrase, bt.current),
|
|
561
|
+
h,
|
|
562
|
+
O
|
|
563
|
+
);
|
|
564
|
+
}
|
|
565
|
+
if (Q.current = !1, Mt && h())
|
|
566
|
+
throw new Error("AudioContext blocked — tap Speak / Continue Learning to unlock audio");
|
|
567
|
+
if (h() && yt) (It = L.current) == null || It.call(L);
|
|
568
|
+
else if (h() && !yt)
|
|
569
|
+
throw new Error("No audio could be played");
|
|
570
|
+
}, []), Ot = K((u) => {
|
|
571
|
+
var T, A, k;
|
|
572
|
+
if (!Pt(n)) {
|
|
573
|
+
(T = D.current) == null || T.call(D, new Error('Voice-only mode requires ttsService="deepgram"'));
|
|
574
|
+
return;
|
|
575
|
+
}
|
|
576
|
+
if (!P.current || !(u != null && u.trim())) return;
|
|
577
|
+
const d = kt(
|
|
578
|
+
n === "deepgram" ? V.current || (S == null ? void 0 : S.VITE_DEEPGRAM_API_KEY) : V.current
|
|
579
|
+
);
|
|
580
|
+
if (n !== "deepgram" || Tt(d, I.current)) {
|
|
581
|
+
const v = Et(u)[0] || u.trim();
|
|
582
|
+
v && ht(
|
|
583
|
+
F.current,
|
|
584
|
+
P.current.audioCtx,
|
|
585
|
+
d,
|
|
586
|
+
it.current || "aura-2-thalia-en",
|
|
587
|
+
v,
|
|
588
|
+
null,
|
|
589
|
+
st.current,
|
|
590
|
+
I.current
|
|
591
|
+
);
|
|
592
|
+
}
|
|
593
|
+
X(!0), j.current = !1, Z.current = !1, O.current = !1;
|
|
594
|
+
const w = J.current;
|
|
595
|
+
Y(P.current.audioCtx), (A = at.current) == null || A.call(at, u), (k = $.current) == null || k.call($, ""), $t(u, w).catch((g) => {
|
|
596
|
+
var v, h;
|
|
597
|
+
J.current !== w || Z.current || (console.error("[VoiceOnlyNarrator] Speech failed:", g), Q.current = !1, (v = L.current) == null || v.call(L), (h = D.current) == null || h.call(D, g));
|
|
598
|
+
});
|
|
599
|
+
}, [X, $t, n]), gt = K(() => {
|
|
600
|
+
var d, w;
|
|
601
|
+
const u = P.current;
|
|
602
|
+
if (u) {
|
|
603
|
+
if (u.isAudioPlaying && G.current.segmentStartedCtxTime != null) {
|
|
604
|
+
const T = (u.audioCtx.currentTime - G.current.segmentStartedCtxTime) * 1e3;
|
|
605
|
+
G.current.resumeOffsetMs = (G.current.segmentOffsetMs || 0) + Math.max(0, T);
|
|
606
|
+
}
|
|
607
|
+
O.current = !0;
|
|
608
|
+
try {
|
|
609
|
+
(w = (d = u.source) == null ? void 0 : d.stop) == null || w.call(d);
|
|
610
|
+
} catch {
|
|
611
|
+
}
|
|
612
|
+
u.isAudioPlaying = !1;
|
|
613
|
+
}
|
|
614
|
+
}, []), pt = K(async () => {
|
|
615
|
+
var u;
|
|
616
|
+
O.current && (await Y((u = P.current) == null ? void 0 : u.audioCtx), O.current = !1);
|
|
617
|
+
}, []), Dt = K(() => {
|
|
618
|
+
var u;
|
|
619
|
+
j.current = !1, X(!0), (u = L.current) == null || u.call(L);
|
|
620
|
+
}, [X]);
|
|
621
|
+
return ct.current = gt, lt.current = pt, qt(b, () => ({
|
|
622
|
+
speakText: Ot,
|
|
623
|
+
preloadSpeech: dt,
|
|
624
|
+
pauseSpeaking: gt,
|
|
625
|
+
resumeSpeaking: pt,
|
|
626
|
+
stopSpeaking: Dt,
|
|
627
|
+
makeEyeContact: () => {
|
|
628
|
+
},
|
|
629
|
+
setMood: () => {
|
|
630
|
+
},
|
|
631
|
+
setLighting: () => {
|
|
632
|
+
},
|
|
633
|
+
setView: () => {
|
|
634
|
+
},
|
|
635
|
+
playGesture: () => {
|
|
636
|
+
},
|
|
637
|
+
playTeachingGesture: () => !1,
|
|
638
|
+
playAnimation: () => {
|
|
639
|
+
},
|
|
640
|
+
playPose: () => {
|
|
641
|
+
},
|
|
642
|
+
stopAnimation: () => {
|
|
643
|
+
},
|
|
644
|
+
stopPose: () => {
|
|
645
|
+
},
|
|
646
|
+
setMixerGain: () => {
|
|
647
|
+
},
|
|
648
|
+
playBackgroundAudio: () => {
|
|
649
|
+
},
|
|
650
|
+
stopBackgroundAudio: () => {
|
|
651
|
+
},
|
|
652
|
+
isReady: U,
|
|
653
|
+
get isSpeaking() {
|
|
654
|
+
var u;
|
|
655
|
+
return Q.current || !!((u = P.current) != null && u.isSpeaking);
|
|
656
|
+
},
|
|
657
|
+
get talkingHead() {
|
|
658
|
+
return null;
|
|
659
|
+
},
|
|
660
|
+
get avatarMode() {
|
|
661
|
+
return "voice";
|
|
662
|
+
}
|
|
663
|
+
}), [Ot, dt, gt, pt, Dt, U]), /* @__PURE__ */ z(
|
|
664
|
+
"div",
|
|
665
|
+
{
|
|
666
|
+
className: `narrator-avatar-container narrator-avatar-voice-only ${nt}`,
|
|
667
|
+
style: {
|
|
668
|
+
position: "relative",
|
|
669
|
+
minHeight: 0,
|
|
670
|
+
height: 0,
|
|
671
|
+
overflow: "hidden",
|
|
672
|
+
pointerEvents: "none",
|
|
673
|
+
...H,
|
|
674
|
+
// Force zero footprint — host should not reserve avatar layout on mobile
|
|
675
|
+
width: (H == null ? void 0 : H.width) ?? "100%",
|
|
676
|
+
maxHeight: 0
|
|
677
|
+
},
|
|
678
|
+
"data-avatar-mode": "voice",
|
|
679
|
+
"aria-hidden": "true"
|
|
680
|
+
}
|
|
681
|
+
);
|
|
682
|
+
});
|
|
683
|
+
_t.displayName = "VoiceOnlyNarrator";
|
|
684
|
+
const me = zt(() => import("./NarratorAvatarFull-DhBLV6kE.js"));
|
|
685
|
+
function we() {
|
|
686
|
+
var r, o, e, n;
|
|
687
|
+
if (typeof window > "u") return !1;
|
|
688
|
+
const t = typeof navigator < "u" && navigator.userAgent || "";
|
|
689
|
+
return !!(/iPhone|iPod|Android.*Mobile|webOS|BlackBerry|IEMobile|Opera Mini/i.test(t) || /iPad/i.test(t) || /Android/i.test(t) && !/Mobile/i.test(t) || /Macintosh/i.test(t) && (navigator.maxTouchPoints || 0) > 1 && ((o = (r = window.matchMedia) == null ? void 0 : r.call(window, "(pointer: coarse)")) != null && o.matches) || (n = (e = window.matchMedia) == null ? void 0 : e.call(window, "(max-width: 768px)")) != null && n.matches || window.innerWidth <= 768);
|
|
690
|
+
}
|
|
691
|
+
function ge(t, r) {
|
|
692
|
+
return t === "voice" || t === !0 ? "voice" : t === "full" || t === !1 ? "full" : we() && r === "deepgram" ? "voice" : "full";
|
|
693
|
+
}
|
|
694
|
+
const pe = Gt(function(r, o) {
|
|
695
|
+
var a;
|
|
696
|
+
const { avatarMode: e = "auto", ...n } = r, [i, s] = Vt(() => e === "voice" || e === !0 ? "voice" : e === "full" || e === !1 ? "full" : null);
|
|
697
|
+
return et(() => {
|
|
698
|
+
s(ge(e, n.ttsService || "google"));
|
|
699
|
+
}, [e, n.ttsService]), i === null ? /* @__PURE__ */ z(
|
|
700
|
+
"div",
|
|
701
|
+
{
|
|
702
|
+
className: `narrator-avatar-container ${n.className || ""}`,
|
|
703
|
+
style: { position: "relative", minHeight: 0, ...n.style },
|
|
704
|
+
"data-avatar-mode": "resolving",
|
|
705
|
+
"aria-hidden": "true"
|
|
706
|
+
}
|
|
707
|
+
) : i === "voice" ? /* @__PURE__ */ z(_t, { ref: o, ...n }) : /* @__PURE__ */ z(
|
|
708
|
+
Yt,
|
|
709
|
+
{
|
|
710
|
+
fallback: /* @__PURE__ */ z(
|
|
711
|
+
"div",
|
|
712
|
+
{
|
|
713
|
+
className: `narrator-avatar-container ${n.className || ""}`,
|
|
714
|
+
style: {
|
|
715
|
+
position: "relative",
|
|
716
|
+
display: "flex",
|
|
717
|
+
alignItems: "center",
|
|
718
|
+
justifyContent: "center",
|
|
719
|
+
minHeight: ((a = n.style) == null ? void 0 : a.minHeight) || 400,
|
|
720
|
+
...n.style
|
|
721
|
+
},
|
|
722
|
+
children: "Loading 3D avatar…"
|
|
723
|
+
}
|
|
724
|
+
),
|
|
725
|
+
children: /* @__PURE__ */ z(me, { ref: o, ...n })
|
|
726
|
+
}
|
|
727
|
+
);
|
|
728
|
+
});
|
|
729
|
+
pe.displayName = "NarratorAvatar";
|
|
730
|
+
export {
|
|
731
|
+
pe as N,
|
|
732
|
+
Tt as c
|
|
733
|
+
};
|
|
734
|
+
//# sourceMappingURL=index-xRyAk0sW.js.map
|