@streamoji/aitwin 0.1.1
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/README.md +102 -0
- package/dist/aitwin.cjs +3 -0
- package/dist/aitwin.js +1779 -0
- package/dist/assets/visemeDiffPreview.worker-B8Juk7ys.js +1 -0
- package/dist/index.d.ts +176 -0
- package/package.json +45 -0
package/dist/aitwin.js
ADDED
|
@@ -0,0 +1,1779 @@
|
|
|
1
|
+
import { jsxs as ft, jsx as We } from "react/jsx-runtime";
|
|
2
|
+
import { forwardRef as dt, useRef as M, useState as _e, useCallback as ee, useEffect as le, useMemo as ht, useImperativeHandle as mt } from "react";
|
|
3
|
+
const $e = "https://ai.streamoji.com", tt = "https://pub-607ad1fc22e2400eb57d17240aab857c.r2.dev/custom-faces", _t = "https://us-central1-streamoji-265f4.cloudfunctions.net/getAiTwin";
|
|
4
|
+
class Pe extends Error {
|
|
5
|
+
constructor(e = "AI twin not found") {
|
|
6
|
+
super(e), this.name = "AiTwinNotFoundError";
|
|
7
|
+
}
|
|
8
|
+
}
|
|
9
|
+
async function wt(t, e = _t) {
|
|
10
|
+
const n = new URL(e);
|
|
11
|
+
n.searchParams.set("id", t);
|
|
12
|
+
const r = await fetch(n.toString());
|
|
13
|
+
if (r.status === 404)
|
|
14
|
+
throw new Pe();
|
|
15
|
+
if (!r.ok)
|
|
16
|
+
throw new Error(`getAiTwin failed (${r.status})`);
|
|
17
|
+
const i = await r.json();
|
|
18
|
+
if (!i.success)
|
|
19
|
+
throw new Pe(i.error ?? "AI twin not found");
|
|
20
|
+
return i.data;
|
|
21
|
+
}
|
|
22
|
+
const pt = "https://us-central1-streamoji-265f4.cloudfunctions.net/getAuthToken", gt = "client_6TNvp3SCs4Og0a1ijm9TommXLql1", yt = "kk8Fq8EmexzP10jMIEwY3R44M5RKUEm1", kt = "6TNvp3SCs4Og0a1ijm9TommXLql1", Et = "Swaraj Mali";
|
|
23
|
+
async function Ke() {
|
|
24
|
+
const t = await fetch(pt, {
|
|
25
|
+
method: "POST",
|
|
26
|
+
headers: {
|
|
27
|
+
"Content-Type": "application/json",
|
|
28
|
+
"Client-Id": gt,
|
|
29
|
+
"Client-Secret": yt
|
|
30
|
+
},
|
|
31
|
+
body: JSON.stringify({
|
|
32
|
+
userId: kt,
|
|
33
|
+
userName: Et
|
|
34
|
+
})
|
|
35
|
+
});
|
|
36
|
+
if (!t.ok)
|
|
37
|
+
throw new Error(`getAuthToken failed (${t.status})`);
|
|
38
|
+
const e = await t.json(), n = e.authToken ?? e.token ?? "";
|
|
39
|
+
if (!n.trim())
|
|
40
|
+
throw new Error("getAuthToken returned no token");
|
|
41
|
+
return n;
|
|
42
|
+
}
|
|
43
|
+
async function St(t, e = $e) {
|
|
44
|
+
const n = await fetch(`${e.replace(/\/$/, "")}/api/session-value`, {
|
|
45
|
+
headers: { Authorization: `Bearer ${t}` }
|
|
46
|
+
});
|
|
47
|
+
if (!n.ok)
|
|
48
|
+
throw new Error(`/api/session-value failed (${n.status})`);
|
|
49
|
+
const r = await n.json();
|
|
50
|
+
if (!r.value) throw new Error("Session value response missing value");
|
|
51
|
+
return r;
|
|
52
|
+
}
|
|
53
|
+
function vt(t) {
|
|
54
|
+
const e = t.trim();
|
|
55
|
+
if (e.length % 2 !== 0) throw new Error("Invalid hex key");
|
|
56
|
+
const n = new Uint8Array(e.length / 2);
|
|
57
|
+
for (let r = 0; r < e.length; r += 2)
|
|
58
|
+
n[r / 2] = Number.parseInt(e.slice(r, r + 2), 16);
|
|
59
|
+
return n;
|
|
60
|
+
}
|
|
61
|
+
function ve(t) {
|
|
62
|
+
return Uint8Array.from(t);
|
|
63
|
+
}
|
|
64
|
+
function At(t, e) {
|
|
65
|
+
const n = new Uint8Array(t.length + e.length);
|
|
66
|
+
return n.set(t, 0), n.set(e, t.length), n;
|
|
67
|
+
}
|
|
68
|
+
async function Tt(t, e) {
|
|
69
|
+
if (!e) throw new Error("Asset key is required for encrypted assets");
|
|
70
|
+
if (t.byteLength < 28)
|
|
71
|
+
throw new Error("Encrypted payload too small");
|
|
72
|
+
const n = new Uint8Array(t), r = ve(n.subarray(0, 12)), i = ve(n.subarray(12, 28)), a = ve(n.subarray(28)), s = ve(At(a, i)), c = await crypto.subtle.importKey(
|
|
73
|
+
"raw",
|
|
74
|
+
vt(e),
|
|
75
|
+
{ name: "AES-GCM" },
|
|
76
|
+
!1,
|
|
77
|
+
["decrypt"]
|
|
78
|
+
);
|
|
79
|
+
return crypto.subtle.decrypt(
|
|
80
|
+
{ name: "AES-GCM", iv: r, tagLength: 128 },
|
|
81
|
+
c,
|
|
82
|
+
s
|
|
83
|
+
);
|
|
84
|
+
}
|
|
85
|
+
let Ce = null;
|
|
86
|
+
const Fe = /* @__PURE__ */ new Set();
|
|
87
|
+
function It(t, e = tt) {
|
|
88
|
+
const n = `${e.replace(/\/$/, "")}/${t}`;
|
|
89
|
+
return { twinBase: n, binBase: n, encrypted: !0 };
|
|
90
|
+
}
|
|
91
|
+
function Rt(t) {
|
|
92
|
+
Ce = t;
|
|
93
|
+
for (const e of Fe) e();
|
|
94
|
+
}
|
|
95
|
+
function re() {
|
|
96
|
+
if (!Ce)
|
|
97
|
+
throw new Error("Twin assets not configured; call setActiveTwinAssets first");
|
|
98
|
+
return Ce;
|
|
99
|
+
}
|
|
100
|
+
function bt(t) {
|
|
101
|
+
return Fe.add(t), () => Fe.delete(t);
|
|
102
|
+
}
|
|
103
|
+
function xt() {
|
|
104
|
+
return `${re().twinBase}/sil.png`;
|
|
105
|
+
}
|
|
106
|
+
function Dt() {
|
|
107
|
+
return `${re().twinBase}/idle.mp4`;
|
|
108
|
+
}
|
|
109
|
+
function Pt() {
|
|
110
|
+
const { encrypted: t, binBase: e, twinBase: n } = re();
|
|
111
|
+
return t ? e : n;
|
|
112
|
+
}
|
|
113
|
+
function ne() {
|
|
114
|
+
return re().encrypted;
|
|
115
|
+
}
|
|
116
|
+
function nt() {
|
|
117
|
+
const { binBase: t, twinBase: e, encrypted: n } = re();
|
|
118
|
+
return n ? `${t}/atlas.bin` : `${e}/atlas.json`;
|
|
119
|
+
}
|
|
120
|
+
function Ct() {
|
|
121
|
+
const { binBase: t, twinBase: e, encrypted: n } = re();
|
|
122
|
+
return n ? `${t}/expression_atlas.bin` : `${e}/expression_atlas.json`;
|
|
123
|
+
}
|
|
124
|
+
function Ft(t) {
|
|
125
|
+
const e = t.split(/\r?\n/);
|
|
126
|
+
let n = "", r = "";
|
|
127
|
+
for (const a of e)
|
|
128
|
+
a.startsWith("event:") ? n = a.slice(6).trim() : a.startsWith("data:") && (r = a.slice(5).trim());
|
|
129
|
+
if (!n) return null;
|
|
130
|
+
let i = {};
|
|
131
|
+
if (r)
|
|
132
|
+
try {
|
|
133
|
+
i = JSON.parse(r);
|
|
134
|
+
} catch {
|
|
135
|
+
i = { raw: r };
|
|
136
|
+
}
|
|
137
|
+
return { event: n, data: i };
|
|
138
|
+
}
|
|
139
|
+
const Mt = [
|
|
140
|
+
"aa",
|
|
141
|
+
"CH",
|
|
142
|
+
"DD",
|
|
143
|
+
"E",
|
|
144
|
+
"FF",
|
|
145
|
+
"I",
|
|
146
|
+
"O",
|
|
147
|
+
"PP",
|
|
148
|
+
"RR",
|
|
149
|
+
"SS",
|
|
150
|
+
"TH",
|
|
151
|
+
"U",
|
|
152
|
+
"kk",
|
|
153
|
+
"nn",
|
|
154
|
+
"sil"
|
|
155
|
+
], Ot = new Map(
|
|
156
|
+
Mt.map((t) => [t.toLowerCase(), t])
|
|
157
|
+
);
|
|
158
|
+
function Ut(t) {
|
|
159
|
+
const e = (t ?? "").trim();
|
|
160
|
+
return e ? Ot.get(e.toLowerCase()) ?? "sil" : "sil";
|
|
161
|
+
}
|
|
162
|
+
const Lt = {
|
|
163
|
+
aei: ["aa", "E", "I"],
|
|
164
|
+
o: ["O", "U"],
|
|
165
|
+
ee: ["I"],
|
|
166
|
+
bmp: ["PP"],
|
|
167
|
+
fv: ["FF"],
|
|
168
|
+
l: ["nn"],
|
|
169
|
+
r: ["RR"],
|
|
170
|
+
th: ["TH"],
|
|
171
|
+
qw: ["U", "O"],
|
|
172
|
+
cdgknstxyz: ["DD", "SS", "kk", "CH"]
|
|
173
|
+
};
|
|
174
|
+
function $t(t) {
|
|
175
|
+
if (!t) return ["sil"];
|
|
176
|
+
const e = t.toLowerCase();
|
|
177
|
+
return Lt[e] ?? ["sil"];
|
|
178
|
+
}
|
|
179
|
+
function Bt(t, e) {
|
|
180
|
+
const n = t.word.length;
|
|
181
|
+
if (n <= 0) return 0;
|
|
182
|
+
const r = t.wduration;
|
|
183
|
+
if (r <= 0) return 0;
|
|
184
|
+
const i = Math.max(0, Math.min(1, (e - t.wtime) / r));
|
|
185
|
+
return Math.min(n - 1, Math.floor(i * n));
|
|
186
|
+
}
|
|
187
|
+
function xe(t, e) {
|
|
188
|
+
const n = Math.max(0, Math.min(t.length - 1, e));
|
|
189
|
+
return t[n] ?? "";
|
|
190
|
+
}
|
|
191
|
+
function Vt(t, e) {
|
|
192
|
+
const n = t.toLowerCase(), r = xe(n, e), i = n.slice(Math.max(0, e - 1), e + 2);
|
|
193
|
+
return r === "i" || r === "y" ? "I" : i.includes("ee") || r === "e" || n.includes("ea") ? "E" : "aa";
|
|
194
|
+
}
|
|
195
|
+
function Ht(t, e) {
|
|
196
|
+
const n = t.toLowerCase(), r = n.slice(
|
|
197
|
+
Math.max(0, e - 1),
|
|
198
|
+
Math.min(n.length, e + 3)
|
|
199
|
+
);
|
|
200
|
+
if (/oo|ou|uw/.test(r)) return "U";
|
|
201
|
+
const i = xe(n, e);
|
|
202
|
+
return i === "u" || i === "w" ? "U" : "O";
|
|
203
|
+
}
|
|
204
|
+
function Nt(t, e) {
|
|
205
|
+
const n = t.toLowerCase(), r = n.slice(Math.max(0, e - 2), e + 1);
|
|
206
|
+
if (/qu/.test(r)) return "U";
|
|
207
|
+
const i = n.slice(
|
|
208
|
+
Math.max(0, e - 1),
|
|
209
|
+
Math.min(n.length, e + 3)
|
|
210
|
+
);
|
|
211
|
+
return /wo|wh/.test(i) || xe(n, e) === "o" ? "O" : "U";
|
|
212
|
+
}
|
|
213
|
+
const Ae = {
|
|
214
|
+
d: "DD",
|
|
215
|
+
t: "DD",
|
|
216
|
+
n: "DD",
|
|
217
|
+
s: "SS",
|
|
218
|
+
z: "SS",
|
|
219
|
+
k: "kk",
|
|
220
|
+
g: "kk",
|
|
221
|
+
c: "CH",
|
|
222
|
+
x: "CH",
|
|
223
|
+
j: "CH"
|
|
224
|
+
};
|
|
225
|
+
function Wt(t, e) {
|
|
226
|
+
const n = t.toLowerCase(), r = xe(n, e);
|
|
227
|
+
if (Ae[r]) return Ae[r];
|
|
228
|
+
for (let i = 0; i < n.length; i++) {
|
|
229
|
+
const a = n[i];
|
|
230
|
+
if (Ae[a]) return Ae[a];
|
|
231
|
+
}
|
|
232
|
+
return "DD";
|
|
233
|
+
}
|
|
234
|
+
function Kt(t, e, n) {
|
|
235
|
+
const r = $t(t);
|
|
236
|
+
if (r.length === 1) return r[0];
|
|
237
|
+
const i = e.trim();
|
|
238
|
+
if (!i) return r[0];
|
|
239
|
+
const a = (t ?? "").toLowerCase(), s = Math.max(0, n);
|
|
240
|
+
switch (a) {
|
|
241
|
+
case "aei":
|
|
242
|
+
return Vt(i, s);
|
|
243
|
+
case "o":
|
|
244
|
+
return Ht(i, s);
|
|
245
|
+
case "qw":
|
|
246
|
+
return Nt(i, s);
|
|
247
|
+
case "cdgknstxyz":
|
|
248
|
+
return Wt(i, s);
|
|
249
|
+
default:
|
|
250
|
+
return r[0];
|
|
251
|
+
}
|
|
252
|
+
}
|
|
253
|
+
function qt(t) {
|
|
254
|
+
let e = t.trim();
|
|
255
|
+
const n = e.indexOf(",");
|
|
256
|
+
e.startsWith("data:") && n >= 0 && (e = e.slice(n + 1)), e = e.replace(/\s+/g, "").replace(/-/g, "+").replace(/_/g, "/");
|
|
257
|
+
const r = e.length % 4;
|
|
258
|
+
return r !== 0 && (e += "=".repeat(4 - r)), e;
|
|
259
|
+
}
|
|
260
|
+
const jt = 24e3;
|
|
261
|
+
function zt(t) {
|
|
262
|
+
return t.length >= 4 && t[0] === 82 && t[1] === 73 && t[2] === 70 && t[3] === 70;
|
|
263
|
+
}
|
|
264
|
+
function qe(t, e, n = jt) {
|
|
265
|
+
const r = e.byteOffset % 2 === 0 ? e : e.slice(), i = new Int16Array(
|
|
266
|
+
r.buffer,
|
|
267
|
+
r.byteOffset,
|
|
268
|
+
r.byteLength / 2
|
|
269
|
+
), a = new Float32Array(i.length);
|
|
270
|
+
for (let c = 0; c < i.length; c++)
|
|
271
|
+
a[c] = i[c] >= 32768 ? -(65536 - i[c]) / 32768 : i[c] / 32767;
|
|
272
|
+
const s = t.createBuffer(1, a.length, n);
|
|
273
|
+
return s.copyToChannel(a, 0), s;
|
|
274
|
+
}
|
|
275
|
+
function Gt(t) {
|
|
276
|
+
const e = qt(t), n = window.atob(e), r = new Uint8Array(n.length);
|
|
277
|
+
for (let i = 0; i < n.length; i++)
|
|
278
|
+
r[i] = n.charCodeAt(i);
|
|
279
|
+
return r;
|
|
280
|
+
}
|
|
281
|
+
async function Qt(t, e) {
|
|
282
|
+
const n = Gt(e);
|
|
283
|
+
if (zt(n)) {
|
|
284
|
+
const r = n.buffer.slice(
|
|
285
|
+
n.byteOffset,
|
|
286
|
+
n.byteOffset + n.byteLength
|
|
287
|
+
);
|
|
288
|
+
return t.decodeAudioData(r);
|
|
289
|
+
}
|
|
290
|
+
if (n.length >= 2 && n.length % 2 === 0)
|
|
291
|
+
return qe(t, n);
|
|
292
|
+
try {
|
|
293
|
+
const r = n.buffer.slice(
|
|
294
|
+
n.byteOffset,
|
|
295
|
+
n.byteOffset + n.byteLength
|
|
296
|
+
);
|
|
297
|
+
return await t.decodeAudioData(r);
|
|
298
|
+
} catch {
|
|
299
|
+
return qe(t, n);
|
|
300
|
+
}
|
|
301
|
+
}
|
|
302
|
+
function Jt(t, e) {
|
|
303
|
+
let n = null;
|
|
304
|
+
for (const r of t)
|
|
305
|
+
e < r.vtime || e >= r.vtime + r.vduration || (!n || r.vtime >= n.vtime) && (n = r);
|
|
306
|
+
return n ? n.viseme : "sil";
|
|
307
|
+
}
|
|
308
|
+
function Xt(t, e) {
|
|
309
|
+
for (const n of t)
|
|
310
|
+
if (e >= n.wtime && e < n.wtime + n.wduration)
|
|
311
|
+
return n;
|
|
312
|
+
return null;
|
|
313
|
+
}
|
|
314
|
+
function je(t) {
|
|
315
|
+
let e = 0;
|
|
316
|
+
for (const n of t)
|
|
317
|
+
e = Math.max(e, n.vtime + n.vduration);
|
|
318
|
+
return e;
|
|
319
|
+
}
|
|
320
|
+
const Yt = 0.5, Zt = 1.5, rt = "Olivia", en = "f786b574-daa5-4673-aa0c-cbe3e8534c02", it = 0.85;
|
|
321
|
+
function tn(t, e = $e) {
|
|
322
|
+
const n = `${e.replace(/\/$/, "")}/avatar_ttsWithPoses`, r = [], i = [], a = [];
|
|
323
|
+
let s = "neutral";
|
|
324
|
+
const c = [], u = [];
|
|
325
|
+
let p = null, _ = !1, o = !1, f = !1, d = !1, k = 0, x = 0, F = 0, Z = 0, H = null, h = 1, g = "", P = "google", O = "google", C = !1;
|
|
326
|
+
const K = /* @__PURE__ */ new Set(), ie = () => {
|
|
327
|
+
for (const m of u)
|
|
328
|
+
m.playbackRate.value = h;
|
|
329
|
+
}, D = () => K.forEach((m) => m()), N = () => {
|
|
330
|
+
H != null && (clearTimeout(H), H = null);
|
|
331
|
+
}, se = () => {
|
|
332
|
+
for (const m of u)
|
|
333
|
+
try {
|
|
334
|
+
m.stop();
|
|
335
|
+
} catch {
|
|
336
|
+
}
|
|
337
|
+
u.length = 0;
|
|
338
|
+
}, fe = () => {
|
|
339
|
+
C = !0, N(), se(), c.length = 0, _ = !1, o = !1, f = !1, d = !1, r.length = 0, i.length = 0, a.length = 0, s = "neutral", F = 0, Z = 0, x = 0, t("idle"), D();
|
|
340
|
+
}, oe = () => {
|
|
341
|
+
!d || _ || u.length > 0 || c.length > 0 || (o = !1, N(), t("done"), D());
|
|
342
|
+
}, de = () => {
|
|
343
|
+
const m = je(r);
|
|
344
|
+
if (m <= 0) {
|
|
345
|
+
t("done"), D();
|
|
346
|
+
return;
|
|
347
|
+
}
|
|
348
|
+
o = !0, k = performance.now(), t("speaking"), D(), N();
|
|
349
|
+
const R = h > 0 ? m / h : m;
|
|
350
|
+
H = setTimeout(() => {
|
|
351
|
+
C || (o = !1, H = null, t("done"), D());
|
|
352
|
+
}, R);
|
|
353
|
+
}, z = (m, R, A) => {
|
|
354
|
+
m.length !== 0 && (A && (F = R), m.forEach((T) => {
|
|
355
|
+
const I = String(T.word ?? "").trim();
|
|
356
|
+
if (!I) return;
|
|
357
|
+
const q = Math.round((T.start ?? 0) * 1e3), B = Math.round((T.duration ?? 0) * 1e3);
|
|
358
|
+
i.push({
|
|
359
|
+
word: I,
|
|
360
|
+
wtime: F + q,
|
|
361
|
+
wduration: B,
|
|
362
|
+
queueIndex: i.length
|
|
363
|
+
});
|
|
364
|
+
}));
|
|
365
|
+
}, ke = (m, R, A, T) => {
|
|
366
|
+
m.length !== 0 && (A && (F = R), m.forEach((b) => {
|
|
367
|
+
const I = String(b.symbol ?? "").trim();
|
|
368
|
+
if (!I) return;
|
|
369
|
+
const q = Math.round((b.start ?? 0) * 1e3), B = Math.max(
|
|
370
|
+
1,
|
|
371
|
+
Math.round((b.duration ?? 0) * 1e3)
|
|
372
|
+
), v = F + q;
|
|
373
|
+
let G;
|
|
374
|
+
if (T === "google" || T === "cartesia")
|
|
375
|
+
G = Ut(I);
|
|
376
|
+
else {
|
|
377
|
+
const S = Xt(i, v), V = S ? Bt(S, v) : 0;
|
|
378
|
+
G = Kt(
|
|
379
|
+
I,
|
|
380
|
+
S?.word ?? "",
|
|
381
|
+
V
|
|
382
|
+
);
|
|
383
|
+
}
|
|
384
|
+
r.push({
|
|
385
|
+
viseme: G,
|
|
386
|
+
weight: 1,
|
|
387
|
+
vtime: v,
|
|
388
|
+
vduration: B
|
|
389
|
+
});
|
|
390
|
+
}), D());
|
|
391
|
+
}, he = (m, R, A, T, b) => {
|
|
392
|
+
z(R, A, T), ke(m, A, T, b);
|
|
393
|
+
}, ae = async (m, R, A, T) => {
|
|
394
|
+
if (C || !m?.trim()) {
|
|
395
|
+
(R.length > 0 || A.length > 0) && !f && he(
|
|
396
|
+
R,
|
|
397
|
+
A,
|
|
398
|
+
F,
|
|
399
|
+
T,
|
|
400
|
+
O
|
|
401
|
+
);
|
|
402
|
+
return;
|
|
403
|
+
}
|
|
404
|
+
if (_) {
|
|
405
|
+
c.push({
|
|
406
|
+
audio: m,
|
|
407
|
+
visemes: R,
|
|
408
|
+
words: A,
|
|
409
|
+
isNewSegment: T
|
|
410
|
+
});
|
|
411
|
+
return;
|
|
412
|
+
}
|
|
413
|
+
_ = !0;
|
|
414
|
+
try {
|
|
415
|
+
const b = window.AudioContext || window.webkitAudioContext, I = p ?? new b();
|
|
416
|
+
I.state === "suspended" && await I.resume(), p = I;
|
|
417
|
+
const q = await Qt(I, m);
|
|
418
|
+
f = !0;
|
|
419
|
+
const B = I.currentTime;
|
|
420
|
+
let v = Z;
|
|
421
|
+
const G = !o;
|
|
422
|
+
v < B && (v = B + 0.1), Z = v + q.duration;
|
|
423
|
+
const S = I.createBufferSource();
|
|
424
|
+
if (S.buffer = q, S.playbackRate.value = h, S.connect(I.destination), u.push(S), C) {
|
|
425
|
+
u.pop();
|
|
426
|
+
return;
|
|
427
|
+
}
|
|
428
|
+
if (G) {
|
|
429
|
+
o = !0, t("speaking"), x = v, F = 0, r.length = 0, i.length = 0;
|
|
430
|
+
const $ = (v - B) * 1e3;
|
|
431
|
+
k = performance.now() + $;
|
|
432
|
+
}
|
|
433
|
+
S.onended = () => {
|
|
434
|
+
const $ = u.indexOf(S);
|
|
435
|
+
$ >= 0 && u.splice($, 1), oe(), D();
|
|
436
|
+
}, S.start(v);
|
|
437
|
+
const V = (v - x) * 1e3;
|
|
438
|
+
he(
|
|
439
|
+
R,
|
|
440
|
+
A,
|
|
441
|
+
V,
|
|
442
|
+
T,
|
|
443
|
+
O
|
|
444
|
+
);
|
|
445
|
+
} catch (b) {
|
|
446
|
+
console.error("[avatarTtsLipsync] audio chunk failed:", b);
|
|
447
|
+
} finally {
|
|
448
|
+
_ = !1;
|
|
449
|
+
const b = c.shift();
|
|
450
|
+
b ? await ae(
|
|
451
|
+
b.audio,
|
|
452
|
+
b.visemes,
|
|
453
|
+
b.words,
|
|
454
|
+
b.isNewSegment
|
|
455
|
+
) : oe();
|
|
456
|
+
}
|
|
457
|
+
}, Ee = () => {
|
|
458
|
+
if (d = !0, !f && r.length > 0) {
|
|
459
|
+
de();
|
|
460
|
+
return;
|
|
461
|
+
}
|
|
462
|
+
oe();
|
|
463
|
+
};
|
|
464
|
+
return {
|
|
465
|
+
setDeveloperToken: (m) => {
|
|
466
|
+
g = m;
|
|
467
|
+
},
|
|
468
|
+
setTtsProvider: (m) => {
|
|
469
|
+
P = m;
|
|
470
|
+
},
|
|
471
|
+
getTtsProvider: () => P,
|
|
472
|
+
speak: async (m, R) => {
|
|
473
|
+
fe(), C = !1, t("loading");
|
|
474
|
+
const A = g.trim();
|
|
475
|
+
if (!A)
|
|
476
|
+
throw new Error("Developer token required for avatar_ttsWithPoses");
|
|
477
|
+
const T = R?.tts ?? P;
|
|
478
|
+
O = T;
|
|
479
|
+
const b = R?.voiceId ?? rt, I = R?.speakingRate ?? it, q = Math.max(
|
|
480
|
+
Yt,
|
|
481
|
+
Math.min(Zt, I)
|
|
482
|
+
), B = {
|
|
483
|
+
user_query: m,
|
|
484
|
+
tts: T,
|
|
485
|
+
speakingRate: q
|
|
486
|
+
};
|
|
487
|
+
(T === "inworld" || T === "cartesia") && (B.voice_id = b);
|
|
488
|
+
try {
|
|
489
|
+
const v = await fetch(n, {
|
|
490
|
+
method: "POST",
|
|
491
|
+
headers: {
|
|
492
|
+
"Content-Type": "application/json",
|
|
493
|
+
Authorization: `Bearer ${A}`
|
|
494
|
+
},
|
|
495
|
+
body: JSON.stringify(B)
|
|
496
|
+
});
|
|
497
|
+
if (!v.ok)
|
|
498
|
+
throw new Error(`avatar_ttsWithPoses failed (${v.status})`);
|
|
499
|
+
const G = v.body;
|
|
500
|
+
if (!G) throw new Error("No response body");
|
|
501
|
+
const S = G.getReader(), V = new TextDecoder();
|
|
502
|
+
let $ = "";
|
|
503
|
+
const J = async (ce) => {
|
|
504
|
+
const U = Ft(ce);
|
|
505
|
+
if (U) {
|
|
506
|
+
if (U.event === "audio") {
|
|
507
|
+
const l = U.data.chunk, y = U.data.visemes ?? [], E = U.data.words ?? [], w = U.data.is_new_segment ?? !1;
|
|
508
|
+
if (l)
|
|
509
|
+
await ae(l, y, E, w);
|
|
510
|
+
else if (y.length > 0 || E.length > 0) {
|
|
511
|
+
const L = w && r.length > 0 ? je(r) : F;
|
|
512
|
+
he(y, E, L, w, T);
|
|
513
|
+
}
|
|
514
|
+
} else if (U.event === "metadata") {
|
|
515
|
+
const l = U.data.mood;
|
|
516
|
+
typeof l == "string" && l.trim() && (s = l.trim());
|
|
517
|
+
const y = U.data.sentence_emotions;
|
|
518
|
+
if (Array.isArray(y)) {
|
|
519
|
+
a.length = 0;
|
|
520
|
+
for (const E of y) {
|
|
521
|
+
if (!E || typeof E != "object") continue;
|
|
522
|
+
const w = E;
|
|
523
|
+
a.push({
|
|
524
|
+
sentence_index: Number(w.sentence_index ?? 0),
|
|
525
|
+
text: String(w.text ?? ""),
|
|
526
|
+
sentiment: typeof w.sentiment == "string" ? w.sentiment : void 0,
|
|
527
|
+
emotion: String(w.emotion ?? s),
|
|
528
|
+
start_word: Number(w.start_word ?? 0),
|
|
529
|
+
end_word: Number(w.end_word ?? 0)
|
|
530
|
+
});
|
|
531
|
+
}
|
|
532
|
+
}
|
|
533
|
+
D();
|
|
534
|
+
} else if (U.event === "error")
|
|
535
|
+
throw new Error(
|
|
536
|
+
String(U.data.message ?? "avatar_ttsWithPoses stream error")
|
|
537
|
+
);
|
|
538
|
+
}
|
|
539
|
+
};
|
|
540
|
+
for (; ; ) {
|
|
541
|
+
const { done: ce, value: U } = await S.read();
|
|
542
|
+
U && ($ += V.decode(U, { stream: !0 }));
|
|
543
|
+
const l = $.split(`
|
|
544
|
+
|
|
545
|
+
`);
|
|
546
|
+
$ = l.pop() ?? "";
|
|
547
|
+
for (const y of l) await J(y);
|
|
548
|
+
if (ce) {
|
|
549
|
+
$.trim() && await J($.trim()), C || Ee();
|
|
550
|
+
break;
|
|
551
|
+
}
|
|
552
|
+
}
|
|
553
|
+
} catch (v) {
|
|
554
|
+
throw console.error("[avatarTtsLipsync]", v), fe(), t("error"), v;
|
|
555
|
+
}
|
|
556
|
+
},
|
|
557
|
+
stop: fe,
|
|
558
|
+
getVisemeQueue: () => r,
|
|
559
|
+
getWordQueue: () => i,
|
|
560
|
+
getSentenceEmotions: () => a,
|
|
561
|
+
getStreamMood: () => s,
|
|
562
|
+
getPlaybackElapsedMs: () => o ? f && p && x > 0 ? Math.max(
|
|
563
|
+
0,
|
|
564
|
+
(p.currentTime - x) * 1e3
|
|
565
|
+
) : Math.max(0, (performance.now() - k) * h) : 0,
|
|
566
|
+
isSpeaking: () => o,
|
|
567
|
+
setPlaybackSpeed: (m) => {
|
|
568
|
+
h = Math.max(0.1, Math.min(1, m)), ie();
|
|
569
|
+
},
|
|
570
|
+
getPlaybackSpeed: () => h,
|
|
571
|
+
subscribe: (m) => (K.add(m), () => K.delete(m))
|
|
572
|
+
};
|
|
573
|
+
}
|
|
574
|
+
function nn(t, e, n) {
|
|
575
|
+
return {
|
|
576
|
+
width: t.crop?.source_width ?? e,
|
|
577
|
+
height: t.crop?.source_height ?? n
|
|
578
|
+
};
|
|
579
|
+
}
|
|
580
|
+
function rn(t, e) {
|
|
581
|
+
const n = t.crop;
|
|
582
|
+
return {
|
|
583
|
+
x: e?.source_x ?? n?.x ?? 0,
|
|
584
|
+
y: e?.source_y ?? n?.y ?? 0,
|
|
585
|
+
width: e?.source_w ?? n?.width ?? t.frame_width ?? 0,
|
|
586
|
+
height: e?.source_h ?? n?.height ?? t.frame_height ?? 0
|
|
587
|
+
};
|
|
588
|
+
}
|
|
589
|
+
function sn(t) {
|
|
590
|
+
return t ? (t.w ?? 0) > 0 && (t.h ?? 0) > 0 : !1;
|
|
591
|
+
}
|
|
592
|
+
async function st(t, e) {
|
|
593
|
+
const n = await fetch(t);
|
|
594
|
+
if (!n.ok) throw new Error(`Failed to load encrypted asset: ${t}`);
|
|
595
|
+
const r = await n.arrayBuffer();
|
|
596
|
+
return Tt(r, e);
|
|
597
|
+
}
|
|
598
|
+
async function ot(t, e) {
|
|
599
|
+
if (!e) throw new Error("Missing asset key for encrypted JSON");
|
|
600
|
+
const n = await st(t, e), r = new TextDecoder().decode(n);
|
|
601
|
+
return JSON.parse(r);
|
|
602
|
+
}
|
|
603
|
+
async function on(t) {
|
|
604
|
+
const e = new Blob([t], { type: "image/webp" }), n = URL.createObjectURL(e);
|
|
605
|
+
return new Promise((r, i) => {
|
|
606
|
+
const a = new Image();
|
|
607
|
+
a.decoding = "async", a.onload = () => {
|
|
608
|
+
URL.revokeObjectURL(n), r(a);
|
|
609
|
+
}, a.onerror = () => {
|
|
610
|
+
URL.revokeObjectURL(n), i(new Error("Failed to decode decrypted image"));
|
|
611
|
+
}, a.src = n;
|
|
612
|
+
});
|
|
613
|
+
}
|
|
614
|
+
function at(t, e) {
|
|
615
|
+
return `${t}/${e}_minus_sil.png`;
|
|
616
|
+
}
|
|
617
|
+
const Be = 15, er = [
|
|
618
|
+
"aa",
|
|
619
|
+
"E",
|
|
620
|
+
"I",
|
|
621
|
+
"O",
|
|
622
|
+
"U",
|
|
623
|
+
"PP",
|
|
624
|
+
"FF",
|
|
625
|
+
"DD",
|
|
626
|
+
"SS",
|
|
627
|
+
"TH",
|
|
628
|
+
"CH",
|
|
629
|
+
"RR",
|
|
630
|
+
"kk",
|
|
631
|
+
"nn"
|
|
632
|
+
], Me = [
|
|
633
|
+
"t01_0.17",
|
|
634
|
+
"t02_0.33",
|
|
635
|
+
"t03_0.50",
|
|
636
|
+
"t04_0.67",
|
|
637
|
+
"t05_0.83"
|
|
638
|
+
], an = Me.length, ze = /* @__PURE__ */ new Set([
|
|
639
|
+
"aa__kk",
|
|
640
|
+
"aa__nn",
|
|
641
|
+
"aa__sil",
|
|
642
|
+
"CH__aa",
|
|
643
|
+
"CH__DD",
|
|
644
|
+
"CH__E",
|
|
645
|
+
"CH__FF",
|
|
646
|
+
"CH__I",
|
|
647
|
+
"CH__kk",
|
|
648
|
+
"CH__nn",
|
|
649
|
+
"CH__O",
|
|
650
|
+
"CH__PP",
|
|
651
|
+
"CH__RR",
|
|
652
|
+
"CH__sil",
|
|
653
|
+
"CH__SS",
|
|
654
|
+
"CH__TH",
|
|
655
|
+
"CH__U",
|
|
656
|
+
"DD__aa",
|
|
657
|
+
"DD__E",
|
|
658
|
+
"DD__FF",
|
|
659
|
+
"DD__I",
|
|
660
|
+
"DD__kk",
|
|
661
|
+
"DD__nn",
|
|
662
|
+
"DD__O",
|
|
663
|
+
"DD__PP",
|
|
664
|
+
"DD__RR",
|
|
665
|
+
"DD__sil",
|
|
666
|
+
"DD__SS",
|
|
667
|
+
"DD__TH",
|
|
668
|
+
"DD__U",
|
|
669
|
+
"E__aa",
|
|
670
|
+
"E__FF",
|
|
671
|
+
"E__I",
|
|
672
|
+
"E__kk",
|
|
673
|
+
"E__nn",
|
|
674
|
+
"E__O",
|
|
675
|
+
"E__PP",
|
|
676
|
+
"E__RR",
|
|
677
|
+
"E__sil",
|
|
678
|
+
"E__SS",
|
|
679
|
+
"E__TH",
|
|
680
|
+
"E__U",
|
|
681
|
+
"FF__aa",
|
|
682
|
+
"FF__I",
|
|
683
|
+
"FF__kk",
|
|
684
|
+
"FF__nn",
|
|
685
|
+
"FF__O",
|
|
686
|
+
"FF__PP",
|
|
687
|
+
"FF__RR",
|
|
688
|
+
"FF__sil",
|
|
689
|
+
"FF__SS",
|
|
690
|
+
"FF__TH",
|
|
691
|
+
"FF__U",
|
|
692
|
+
"I__aa",
|
|
693
|
+
"I__kk",
|
|
694
|
+
"I__nn",
|
|
695
|
+
"I__O",
|
|
696
|
+
"I__PP",
|
|
697
|
+
"I__RR",
|
|
698
|
+
"I__sil",
|
|
699
|
+
"I__SS",
|
|
700
|
+
"I__TH",
|
|
701
|
+
"I__U",
|
|
702
|
+
"kk__nn",
|
|
703
|
+
"kk__sil",
|
|
704
|
+
"nn__sil",
|
|
705
|
+
"O__aa",
|
|
706
|
+
"O__kk",
|
|
707
|
+
"O__nn",
|
|
708
|
+
"O__PP",
|
|
709
|
+
"O__RR",
|
|
710
|
+
"O__sil",
|
|
711
|
+
"O__SS",
|
|
712
|
+
"O__TH",
|
|
713
|
+
"O__U",
|
|
714
|
+
"PP__aa",
|
|
715
|
+
"PP__kk",
|
|
716
|
+
"PP__nn",
|
|
717
|
+
"PP__RR",
|
|
718
|
+
"PP__sil",
|
|
719
|
+
"PP__SS",
|
|
720
|
+
"PP__TH",
|
|
721
|
+
"PP__U",
|
|
722
|
+
"RR__aa",
|
|
723
|
+
"RR__kk",
|
|
724
|
+
"RR__nn",
|
|
725
|
+
"RR__sil",
|
|
726
|
+
"RR__SS",
|
|
727
|
+
"RR__TH",
|
|
728
|
+
"RR__U",
|
|
729
|
+
"SS__aa",
|
|
730
|
+
"SS__kk",
|
|
731
|
+
"SS__nn",
|
|
732
|
+
"SS__sil",
|
|
733
|
+
"SS__TH",
|
|
734
|
+
"SS__U",
|
|
735
|
+
"TH__aa",
|
|
736
|
+
"TH__kk",
|
|
737
|
+
"TH__nn",
|
|
738
|
+
"TH__sil",
|
|
739
|
+
"TH__U",
|
|
740
|
+
"U__aa",
|
|
741
|
+
"U__kk",
|
|
742
|
+
"U__nn",
|
|
743
|
+
"U__sil"
|
|
744
|
+
]);
|
|
745
|
+
function Ge(t, e) {
|
|
746
|
+
return `${t}__${e}`;
|
|
747
|
+
}
|
|
748
|
+
function Qe(t, e) {
|
|
749
|
+
return `pairs/${t}/${e}_minus_sil.png`;
|
|
750
|
+
}
|
|
751
|
+
function Oe(t) {
|
|
752
|
+
return `${t}_minus_sil.png`;
|
|
753
|
+
}
|
|
754
|
+
function cn(t, e) {
|
|
755
|
+
if (t === e) return [];
|
|
756
|
+
const n = [], r = Ge(t, e);
|
|
757
|
+
if (ze.has(r)) {
|
|
758
|
+
for (const a of Me)
|
|
759
|
+
n.push(Qe(r, a));
|
|
760
|
+
return n;
|
|
761
|
+
}
|
|
762
|
+
const i = Ge(e, t);
|
|
763
|
+
if (ze.has(i)) {
|
|
764
|
+
for (let a = an - 1; a >= 0; a--)
|
|
765
|
+
n.push(Qe(i, Me[a]));
|
|
766
|
+
return n;
|
|
767
|
+
}
|
|
768
|
+
return [];
|
|
769
|
+
}
|
|
770
|
+
function un(t) {
|
|
771
|
+
return t < 16 ? [] : t < 35 ? [2] : t < 51 ? [0, 4] : t < 67 ? [0, 2, 4] : t < 83 ? [0, 1, 2, 3] : [0, 1, 2, 3, 4];
|
|
772
|
+
}
|
|
773
|
+
function ln(t, e) {
|
|
774
|
+
return un(e).map((r) => t[r]).filter((r) => r != null);
|
|
775
|
+
}
|
|
776
|
+
function fn(t, e, n, r) {
|
|
777
|
+
if (r <= 0) return 0;
|
|
778
|
+
const a = Math.max(1, n - e) / r, s = t - e, c = Math.floor(s / a);
|
|
779
|
+
return Math.min(r - 1, Math.max(0, c));
|
|
780
|
+
}
|
|
781
|
+
function dn(t, e) {
|
|
782
|
+
if (!ne()) return `${t}/${e}`;
|
|
783
|
+
const n = e.split("/").pop() ?? e, r = n.includes(".") ? n.slice(0, n.lastIndexOf(".")) : n;
|
|
784
|
+
return `${t}/${r}.bin`;
|
|
785
|
+
}
|
|
786
|
+
async function hn(t) {
|
|
787
|
+
const e = nt(), n = ne() ? await ot(e, t) : await fetch(e).then((r) => {
|
|
788
|
+
if (!r.ok) throw new Error(`Failed to load ${e}`);
|
|
789
|
+
return r.json();
|
|
790
|
+
});
|
|
791
|
+
return {
|
|
792
|
+
width: n.crop?.source_width ?? n.frame_width ?? 842,
|
|
793
|
+
height: n.crop?.source_height ?? n.frame_height ?? 1264
|
|
794
|
+
};
|
|
795
|
+
}
|
|
796
|
+
const mn = [0.32, 0.24, 0.16, 0.08, 0], De = mn.map(
|
|
797
|
+
(t) => `eyes_${t.toFixed(2)}`
|
|
798
|
+
), _n = [
|
|
799
|
+
"eyes_0.32",
|
|
800
|
+
"eyes_0.16",
|
|
801
|
+
"eyes_0.00"
|
|
802
|
+
], wn = [
|
|
803
|
+
"eyes_0.32",
|
|
804
|
+
"eyes_0.16",
|
|
805
|
+
"eyes_0.08",
|
|
806
|
+
"eyes_0.00"
|
|
807
|
+
], pn = {
|
|
808
|
+
fear: "fear",
|
|
809
|
+
afraid: "fear",
|
|
810
|
+
anxiety: "fear",
|
|
811
|
+
anger: "anger",
|
|
812
|
+
angry: "anger",
|
|
813
|
+
mad: "anger",
|
|
814
|
+
sad: "sad",
|
|
815
|
+
sadness: "sad",
|
|
816
|
+
sorrow: "sad",
|
|
817
|
+
neutral: "neutral",
|
|
818
|
+
calm: "neutral",
|
|
819
|
+
love: "love",
|
|
820
|
+
loving: "love",
|
|
821
|
+
affection: "love",
|
|
822
|
+
happy: "happy",
|
|
823
|
+
happiness: "happy",
|
|
824
|
+
joy: "happy",
|
|
825
|
+
joyful: "happy",
|
|
826
|
+
excited: "happy",
|
|
827
|
+
disgust: "disgust",
|
|
828
|
+
disgusted: "disgust"
|
|
829
|
+
};
|
|
830
|
+
function te(t, e = "neutral") {
|
|
831
|
+
const n = String(t ?? "").trim().toLowerCase();
|
|
832
|
+
return n ? pn[n] ?? e : e;
|
|
833
|
+
}
|
|
834
|
+
function W(t) {
|
|
835
|
+
return t === "neutral" ? null : at(t, "eyes_standard");
|
|
836
|
+
}
|
|
837
|
+
function Ue(t, e) {
|
|
838
|
+
return !e || e === "eyes_standard" ? W(t) : t === "neutral" && e === "eyes_standard" ? null : at(t, e);
|
|
839
|
+
}
|
|
840
|
+
function gn(t) {
|
|
841
|
+
if (!t) return "open";
|
|
842
|
+
const e = t.match(/\/(eyes_[^/]+)_minus_sil\.png$/);
|
|
843
|
+
if (!e?.[1]) return "open";
|
|
844
|
+
const n = e[1];
|
|
845
|
+
return De.includes(n) ? n : "open";
|
|
846
|
+
}
|
|
847
|
+
function yn(t) {
|
|
848
|
+
return t === "open" ? -1 : De.indexOf(t);
|
|
849
|
+
}
|
|
850
|
+
function kn(t, e) {
|
|
851
|
+
const n = [];
|
|
852
|
+
for (const r of e)
|
|
853
|
+
n.push(Ue(t, r));
|
|
854
|
+
for (let r = e.length - 2; r >= 0; r--)
|
|
855
|
+
n.push(Ue(t, e[r]));
|
|
856
|
+
return n.push(W(t)), n;
|
|
857
|
+
}
|
|
858
|
+
function En(t, e) {
|
|
859
|
+
if (e === "open") return [W(t)];
|
|
860
|
+
const n = yn(e);
|
|
861
|
+
if (n < 0) return [W(t)];
|
|
862
|
+
const r = [];
|
|
863
|
+
for (let i = n; i >= 0; i--)
|
|
864
|
+
r.push(Ue(t, De[i]));
|
|
865
|
+
return r.push(W(t)), r;
|
|
866
|
+
}
|
|
867
|
+
const Sn = [
|
|
868
|
+
"t01_0.20",
|
|
869
|
+
"t02_0.40",
|
|
870
|
+
"t03_0.60",
|
|
871
|
+
"t04_0.80"
|
|
872
|
+
];
|
|
873
|
+
function vn(t, e) {
|
|
874
|
+
return `${t}__${e}`;
|
|
875
|
+
}
|
|
876
|
+
function An(t, e, n) {
|
|
877
|
+
return `pairs/${vn(t, e)}/${n}_minus_sil.png`;
|
|
878
|
+
}
|
|
879
|
+
function Tn(t, e) {
|
|
880
|
+
return Sn.map(
|
|
881
|
+
(n) => An(t, e, n)
|
|
882
|
+
);
|
|
883
|
+
}
|
|
884
|
+
function pe(t, e) {
|
|
885
|
+
return t + Math.random() * (e - t);
|
|
886
|
+
}
|
|
887
|
+
function Te() {
|
|
888
|
+
return pe(14, 32);
|
|
889
|
+
}
|
|
890
|
+
function In(t = {}) {
|
|
891
|
+
const {
|
|
892
|
+
initialDelayMs: e = 800 + Math.random() * 1200,
|
|
893
|
+
intervalMinMs: n = 2200,
|
|
894
|
+
intervalMaxMs: r = 5800,
|
|
895
|
+
doubleBlinkChance: i = 0.18,
|
|
896
|
+
doubleBlinkGapMinMs: a = 120,
|
|
897
|
+
doubleBlinkGapMaxMs: s = 220,
|
|
898
|
+
fullBlinkChance: c = 0.75,
|
|
899
|
+
postEmotionChangeDelayMs: u = 380
|
|
900
|
+
} = t;
|
|
901
|
+
let p = "neutral", _ = "neutral", o = {
|
|
902
|
+
kind: "idle",
|
|
903
|
+
nextBlinkAt: performance.now() + e
|
|
904
|
+
}, f = null, d = "__eyes_open__|neutral";
|
|
905
|
+
function k(h, g) {
|
|
906
|
+
f = h, d = `${h ?? "__eyes_open__"}|${g}`;
|
|
907
|
+
}
|
|
908
|
+
function x(h, g) {
|
|
909
|
+
o = {
|
|
910
|
+
kind: "idle",
|
|
911
|
+
nextBlinkAt: h + pe(n, r)
|
|
912
|
+
}, _ = g, k(W(g), g);
|
|
913
|
+
}
|
|
914
|
+
function F(h, g, P) {
|
|
915
|
+
const O = P ? _n : Math.random() < c ? De : wn;
|
|
916
|
+
o = {
|
|
917
|
+
kind: "playing",
|
|
918
|
+
paths: kn(g, O),
|
|
919
|
+
index: 0,
|
|
920
|
+
holdUntil: h + Te(),
|
|
921
|
+
after: "idle",
|
|
922
|
+
blinkEmotion: g
|
|
923
|
+
}, _ = g, k(o.paths[0] ?? null, g);
|
|
924
|
+
}
|
|
925
|
+
function Z(h, g, P) {
|
|
926
|
+
const O = gn(f), C = [];
|
|
927
|
+
if (O !== "open") {
|
|
928
|
+
const K = En(g, O);
|
|
929
|
+
C.push(...K.slice(0, -1));
|
|
930
|
+
}
|
|
931
|
+
C.push(...Tn(g, P)), C.push(W(P)), o = {
|
|
932
|
+
kind: "emotionBlend",
|
|
933
|
+
paths: C,
|
|
934
|
+
index: 0,
|
|
935
|
+
holdUntil: h + Te(),
|
|
936
|
+
thenEmotion: P
|
|
937
|
+
}, k(C[0] ?? null, g);
|
|
938
|
+
}
|
|
939
|
+
function H(h) {
|
|
940
|
+
if (o.kind === "emotionBlend") {
|
|
941
|
+
for (; h >= o.holdUntil && o.index < o.paths.length - 1; )
|
|
942
|
+
o.index += 1, o.holdUntil = h + Te();
|
|
943
|
+
const g = o.paths[o.index] ?? null;
|
|
944
|
+
k(g, o.thenEmotion), h >= o.holdUntil && o.index >= o.paths.length - 1 && (p = o.thenEmotion, _ = o.thenEmotion, k(W(o.thenEmotion), o.thenEmotion), o = {
|
|
945
|
+
kind: "idle",
|
|
946
|
+
nextBlinkAt: h + u
|
|
947
|
+
});
|
|
948
|
+
return;
|
|
949
|
+
}
|
|
950
|
+
if (o.kind === "idle") {
|
|
951
|
+
k(W(_), _), h >= o.nextBlinkAt && F(h, _, !1);
|
|
952
|
+
return;
|
|
953
|
+
}
|
|
954
|
+
if (o.kind === "doublePause") {
|
|
955
|
+
k(W(o.blinkEmotion), o.blinkEmotion), h >= o.resumeAt && F(h, o.blinkEmotion, !0);
|
|
956
|
+
return;
|
|
957
|
+
}
|
|
958
|
+
if (o.kind === "playing") {
|
|
959
|
+
for (; h >= o.holdUntil && o.index < o.paths.length - 1; )
|
|
960
|
+
o.index += 1, o.holdUntil = h + Te();
|
|
961
|
+
if (k(o.paths[o.index] ?? null, o.blinkEmotion), h >= o.holdUntil && o.index >= o.paths.length - 1) {
|
|
962
|
+
if (o.after === "resumeIdle") {
|
|
963
|
+
x(h + u, p);
|
|
964
|
+
return;
|
|
965
|
+
}
|
|
966
|
+
if (o.after === "doublePause") {
|
|
967
|
+
o = {
|
|
968
|
+
kind: "doublePause",
|
|
969
|
+
resumeAt: h + pe(a, s),
|
|
970
|
+
blinkEmotion: o.blinkEmotion
|
|
971
|
+
}, k(W(o.blinkEmotion), o.blinkEmotion);
|
|
972
|
+
return;
|
|
973
|
+
}
|
|
974
|
+
if (o.after === "idle" && Math.random() < i) {
|
|
975
|
+
o = {
|
|
976
|
+
kind: "doublePause",
|
|
977
|
+
resumeAt: h + pe(a, s),
|
|
978
|
+
blinkEmotion: o.blinkEmotion
|
|
979
|
+
}, k(W(o.blinkEmotion), o.blinkEmotion);
|
|
980
|
+
return;
|
|
981
|
+
}
|
|
982
|
+
x(h, o.blinkEmotion);
|
|
983
|
+
}
|
|
984
|
+
}
|
|
985
|
+
}
|
|
986
|
+
return {
|
|
987
|
+
advance: H,
|
|
988
|
+
setTargetEmotion(h) {
|
|
989
|
+
const g = te(h, p);
|
|
990
|
+
if (g === p && o.kind !== "emotionBlend") return;
|
|
991
|
+
const P = performance.now(), O = p;
|
|
992
|
+
if (p = g, o.kind === "playing") {
|
|
993
|
+
Z(P, _, g);
|
|
994
|
+
return;
|
|
995
|
+
}
|
|
996
|
+
if (o.kind === "emotionBlend") {
|
|
997
|
+
o = {
|
|
998
|
+
...o,
|
|
999
|
+
thenEmotion: g
|
|
1000
|
+
};
|
|
1001
|
+
return;
|
|
1002
|
+
}
|
|
1003
|
+
Z(P, O, g);
|
|
1004
|
+
},
|
|
1005
|
+
getTargetEmotion: () => p,
|
|
1006
|
+
getExpressionPath: () => f,
|
|
1007
|
+
getDrawKey: () => d,
|
|
1008
|
+
reset() {
|
|
1009
|
+
p = "neutral", _ = "neutral", x(
|
|
1010
|
+
performance.now() + pe(n, r),
|
|
1011
|
+
"neutral"
|
|
1012
|
+
);
|
|
1013
|
+
}
|
|
1014
|
+
};
|
|
1015
|
+
}
|
|
1016
|
+
function Rn(t) {
|
|
1017
|
+
const e = document.createElement("video");
|
|
1018
|
+
e.muted = !0, e.playsInline = !0, e.preload = "auto", e.setAttribute("playsinline", ""), e.loop = !0;
|
|
1019
|
+
let n = !1;
|
|
1020
|
+
return {
|
|
1021
|
+
async load() {
|
|
1022
|
+
n || await new Promise((r, i) => {
|
|
1023
|
+
const a = () => {
|
|
1024
|
+
const s = e.duration;
|
|
1025
|
+
if (!Number.isFinite(s) || s <= 0) {
|
|
1026
|
+
i(new Error(`Idle video has invalid duration: ${t}`));
|
|
1027
|
+
return;
|
|
1028
|
+
}
|
|
1029
|
+
n = !0, e.currentTime = 0, r();
|
|
1030
|
+
};
|
|
1031
|
+
e.addEventListener("loadeddata", a, { once: !0 }), e.addEventListener(
|
|
1032
|
+
"error",
|
|
1033
|
+
() => i(new Error(`Failed to load idle video: ${t}`)),
|
|
1034
|
+
{ once: !0 }
|
|
1035
|
+
), e.src = t, e.load();
|
|
1036
|
+
});
|
|
1037
|
+
},
|
|
1038
|
+
isReady() {
|
|
1039
|
+
return n;
|
|
1040
|
+
},
|
|
1041
|
+
getVideo() {
|
|
1042
|
+
return e;
|
|
1043
|
+
},
|
|
1044
|
+
setActive(r) {
|
|
1045
|
+
n && (r ? e.paused && e.play().catch(() => {
|
|
1046
|
+
}) : e.paused || e.pause());
|
|
1047
|
+
},
|
|
1048
|
+
restart() {
|
|
1049
|
+
n && (e.currentTime = 0);
|
|
1050
|
+
}
|
|
1051
|
+
};
|
|
1052
|
+
}
|
|
1053
|
+
function bn(t, e) {
|
|
1054
|
+
for (const n of t)
|
|
1055
|
+
if (e >= n.wtime && e < n.wtime + n.wduration)
|
|
1056
|
+
return n;
|
|
1057
|
+
return null;
|
|
1058
|
+
}
|
|
1059
|
+
function xn(t, e, n, r = "neutral") {
|
|
1060
|
+
const i = te(r);
|
|
1061
|
+
if (e.length === 0) return i;
|
|
1062
|
+
if (n.length === 0)
|
|
1063
|
+
return te(e[0]?.emotion, i);
|
|
1064
|
+
const a = bn(n, t);
|
|
1065
|
+
let s;
|
|
1066
|
+
if (a)
|
|
1067
|
+
s = a.queueIndex ?? n.findIndex((c) => c === a), s < 0 && (s = 0);
|
|
1068
|
+
else {
|
|
1069
|
+
if (t < n[0].wtime)
|
|
1070
|
+
return te(e[0]?.emotion, i);
|
|
1071
|
+
s = n[n.length - 1].queueIndex ?? n.length - 1;
|
|
1072
|
+
}
|
|
1073
|
+
for (const c of e)
|
|
1074
|
+
if (s >= c.start_word && s <= c.end_word)
|
|
1075
|
+
return te(c.emotion, i);
|
|
1076
|
+
for (const c of e)
|
|
1077
|
+
if (s < c.start_word)
|
|
1078
|
+
return te(c.emotion, i);
|
|
1079
|
+
return te(
|
|
1080
|
+
e[e.length - 1]?.emotion,
|
|
1081
|
+
i
|
|
1082
|
+
);
|
|
1083
|
+
}
|
|
1084
|
+
function ct(t, e, n, r, i = Be) {
|
|
1085
|
+
const a = new OffscreenCanvas(n, r), s = a.getContext("2d", { willReadFrequently: !0 });
|
|
1086
|
+
s.drawImage(t, 0, 0, n, r);
|
|
1087
|
+
const c = s.getImageData(0, 0, n, r), u = s.createImageData(n, r);
|
|
1088
|
+
u.data.set(c.data);
|
|
1089
|
+
const _ = new OffscreenCanvas(n, r).getContext("2d", { willReadFrequently: !0 });
|
|
1090
|
+
for (const o of e) {
|
|
1091
|
+
_.clearRect(0, 0, n, r), _.drawImage(o, 0, 0, n, r);
|
|
1092
|
+
const f = _.getImageData(0, 0, n, r);
|
|
1093
|
+
for (let d = 0; d < u.data.length; d += 4) {
|
|
1094
|
+
const k = f.data[d], x = f.data[d + 1], F = f.data[d + 2];
|
|
1095
|
+
Math.max(k, x, F) > i && (u.data[d] = k, u.data[d + 1] = x, u.data[d + 2] = F, u.data[d + 3] = 255);
|
|
1096
|
+
}
|
|
1097
|
+
}
|
|
1098
|
+
return s.putImageData(u, 0, 0), a;
|
|
1099
|
+
}
|
|
1100
|
+
let ge = null, ye = null;
|
|
1101
|
+
const Re = /* @__PURE__ */ new Map();
|
|
1102
|
+
function Dn() {
|
|
1103
|
+
ge = null, ye = null, Re.clear();
|
|
1104
|
+
}
|
|
1105
|
+
bt(Dn);
|
|
1106
|
+
function Pn(t) {
|
|
1107
|
+
return t ?? ye ?? void 0;
|
|
1108
|
+
}
|
|
1109
|
+
function Ve(t) {
|
|
1110
|
+
return new Promise((e, n) => {
|
|
1111
|
+
const r = new Image();
|
|
1112
|
+
r.decoding = "async", r.onload = () => e(r), r.onerror = () => n(new Error(`Failed to load image: ${t}`)), r.src = t;
|
|
1113
|
+
});
|
|
1114
|
+
}
|
|
1115
|
+
async function Je(t, e, n) {
|
|
1116
|
+
const r = ne() ? await ot(t, n) : await fetch(t).then((u) => {
|
|
1117
|
+
if (!u.ok) throw new Error(`Failed to load ${t}`);
|
|
1118
|
+
return u.json();
|
|
1119
|
+
}), i = r.sheets?.[0];
|
|
1120
|
+
if (!i?.path) throw new Error(`${t} has no sheets[0].path`);
|
|
1121
|
+
const a = ne() ? await on(
|
|
1122
|
+
await st(dn(e, i.path), n ?? "")
|
|
1123
|
+
) : await Ve(`${e}/${i.path}`), s = /* @__PURE__ */ new Map();
|
|
1124
|
+
for (const u of r.cells ?? [])
|
|
1125
|
+
u.path && s.set(u.path, u);
|
|
1126
|
+
const c = /* @__PURE__ */ new Map();
|
|
1127
|
+
for (const u of r.sheets ?? [])
|
|
1128
|
+
c.set(u.index, u);
|
|
1129
|
+
return { atlas: a, atlasMeta: r, cellByPath: s, sheetByIndex: c, diffBase: e };
|
|
1130
|
+
}
|
|
1131
|
+
async function be(t) {
|
|
1132
|
+
const e = Pn(t);
|
|
1133
|
+
if (ne() && !e)
|
|
1134
|
+
throw new Error("Encrypted assets enabled but no key provided");
|
|
1135
|
+
if (ge && (!ne() || ye === (e ?? null)))
|
|
1136
|
+
return ge;
|
|
1137
|
+
ye = e ?? null, Re.clear();
|
|
1138
|
+
const n = Pt();
|
|
1139
|
+
return ge = (async () => {
|
|
1140
|
+
const r = await Ve(xt()), i = await Je(
|
|
1141
|
+
nt(),
|
|
1142
|
+
n,
|
|
1143
|
+
e
|
|
1144
|
+
);
|
|
1145
|
+
let a = null;
|
|
1146
|
+
try {
|
|
1147
|
+
a = await Je(
|
|
1148
|
+
Ct(),
|
|
1149
|
+
n,
|
|
1150
|
+
e
|
|
1151
|
+
);
|
|
1152
|
+
} catch {
|
|
1153
|
+
a = null;
|
|
1154
|
+
}
|
|
1155
|
+
return { sil: r, viseme: i, expression: a };
|
|
1156
|
+
})(), ge;
|
|
1157
|
+
}
|
|
1158
|
+
function He(t) {
|
|
1159
|
+
return nn(
|
|
1160
|
+
t.viseme.atlasMeta,
|
|
1161
|
+
t.sil.naturalWidth,
|
|
1162
|
+
t.sil.naturalHeight
|
|
1163
|
+
);
|
|
1164
|
+
}
|
|
1165
|
+
function Cn(t, e) {
|
|
1166
|
+
const n = document.createElement("canvas");
|
|
1167
|
+
return n.width = e.w, n.height = e.h, n.getContext("2d", { willReadFrequently: !0 }).drawImage(t.atlas, e.x, e.y, e.w, e.h, 0, 0, e.w, e.h), n;
|
|
1168
|
+
}
|
|
1169
|
+
async function Xe(t, e, n) {
|
|
1170
|
+
const r = `${e.diffBase}::${n}`, i = Re.get(r);
|
|
1171
|
+
if (i) return i;
|
|
1172
|
+
const a = (async () => {
|
|
1173
|
+
const s = e.cellByPath.get(n);
|
|
1174
|
+
if (!s) throw new Error(`No atlas cell for path: ${n}`);
|
|
1175
|
+
const c = rn(e.atlasMeta, s), u = He(t);
|
|
1176
|
+
if (sn(s)) {
|
|
1177
|
+
const o = Cn(e, s), f = document.createElement("canvas");
|
|
1178
|
+
f.width = u.width, f.height = u.height;
|
|
1179
|
+
const d = f.getContext("2d");
|
|
1180
|
+
return d.fillStyle = "#000", d.fillRect(0, 0, u.width, u.height), d.drawImage(o, c.x, c.y), f;
|
|
1181
|
+
}
|
|
1182
|
+
const p = `${e.diffBase}/${n}`, _ = await Ve(p);
|
|
1183
|
+
if (_.naturalWidth === u.width && _.naturalHeight === u.height)
|
|
1184
|
+
return _;
|
|
1185
|
+
throw new Error(`Diff PNG wrong size for ${n}`);
|
|
1186
|
+
})();
|
|
1187
|
+
return Re.set(r, a), a;
|
|
1188
|
+
}
|
|
1189
|
+
async function Ye(t, e, n, r = Be) {
|
|
1190
|
+
const i = await be(), { width: a, height: s } = He(i);
|
|
1191
|
+
(t.width !== a || t.height !== s) && (t.width = a, t.height = s);
|
|
1192
|
+
const c = [], u = t.getContext("2d");
|
|
1193
|
+
if (u.clearRect(0, 0, a, s), e.readyState < HTMLMediaElement.HAVE_CURRENT_DATA) {
|
|
1194
|
+
const _ = i.sil;
|
|
1195
|
+
u.drawImage(_, 0, 0, a, s);
|
|
1196
|
+
return;
|
|
1197
|
+
}
|
|
1198
|
+
if (c.length === 0) {
|
|
1199
|
+
u.drawImage(e, 0, 0, a, s);
|
|
1200
|
+
return;
|
|
1201
|
+
}
|
|
1202
|
+
const p = ct(e, c, a, s, r);
|
|
1203
|
+
u.drawImage(p, 0, 0);
|
|
1204
|
+
}
|
|
1205
|
+
async function Le(t, e, n = Be) {
|
|
1206
|
+
const r = await be(ye ?? void 0), { width: i, height: a } = He(r);
|
|
1207
|
+
(t.width !== i || t.height !== a) && (t.width = i, t.height = a);
|
|
1208
|
+
const s = [];
|
|
1209
|
+
e.mouthPath && s.push(await Xe(r, r.viseme, e.mouthPath)), e.expressionPath && r.expression && s.push(
|
|
1210
|
+
await Xe(r, r.expression, e.expressionPath)
|
|
1211
|
+
);
|
|
1212
|
+
const c = t.getContext("2d");
|
|
1213
|
+
if (c.clearRect(0, 0, i, a), s.length === 0) {
|
|
1214
|
+
c.drawImage(r.sil, 0, 0, i, a);
|
|
1215
|
+
return;
|
|
1216
|
+
}
|
|
1217
|
+
const u = ct(
|
|
1218
|
+
r.sil,
|
|
1219
|
+
s,
|
|
1220
|
+
i,
|
|
1221
|
+
a,
|
|
1222
|
+
n
|
|
1223
|
+
);
|
|
1224
|
+
c.drawImage(u, 0, 0);
|
|
1225
|
+
}
|
|
1226
|
+
async function Fn(t) {
|
|
1227
|
+
await Le(t, { mouthPath: null, expressionPath: null });
|
|
1228
|
+
}
|
|
1229
|
+
function Mn(t) {
|
|
1230
|
+
const e = /* @__PURE__ */ new Map();
|
|
1231
|
+
for (const n of t)
|
|
1232
|
+
e.has(n.vtime) || e.set(n.vtime, n);
|
|
1233
|
+
return [...e.values()].sort((n, r) => n.vtime - r.vtime);
|
|
1234
|
+
}
|
|
1235
|
+
function On(t) {
|
|
1236
|
+
const e = Mn(t), n = [];
|
|
1237
|
+
for (let r = 0; r < e.length - 1; r++) {
|
|
1238
|
+
const i = e[r].viseme, a = e[r + 1].viseme, s = e[r].vtime, c = e[r + 1].vtime;
|
|
1239
|
+
i === a || c <= s || n.push({ from: i, to: a, fromVtime: s, toVtime: c });
|
|
1240
|
+
}
|
|
1241
|
+
return n;
|
|
1242
|
+
}
|
|
1243
|
+
function Un(t, e) {
|
|
1244
|
+
let n = null;
|
|
1245
|
+
for (const r of t)
|
|
1246
|
+
e >= r.fromVtime && e < r.toVtime && (n = r);
|
|
1247
|
+
return n;
|
|
1248
|
+
}
|
|
1249
|
+
function Ln(t, e) {
|
|
1250
|
+
return { gapMs: Math.max(0, e - t), holdEnd: t, transStart: t };
|
|
1251
|
+
}
|
|
1252
|
+
const $n = 500, Bn = "__sil__";
|
|
1253
|
+
function Vn() {
|
|
1254
|
+
return { lastDrawnKey: "" };
|
|
1255
|
+
}
|
|
1256
|
+
function we(t) {
|
|
1257
|
+
t.lastDrawnKey = "";
|
|
1258
|
+
}
|
|
1259
|
+
function Ze(t) {
|
|
1260
|
+
return t ?? Bn;
|
|
1261
|
+
}
|
|
1262
|
+
function Hn(t, e) {
|
|
1263
|
+
if (t.length === 0)
|
|
1264
|
+
return { diffPath: null, label: "sil" };
|
|
1265
|
+
const n = On(t), r = Un(n, e);
|
|
1266
|
+
if (r) {
|
|
1267
|
+
const a = r.toVtime - r.fromVtime, s = Ln(r.fromVtime, r.toVtime);
|
|
1268
|
+
if (e >= s.transStart && e < r.toVtime) {
|
|
1269
|
+
const c = cn(r.from, r.to), u = ln(c, a);
|
|
1270
|
+
if (u.length > 0) {
|
|
1271
|
+
const p = fn(
|
|
1272
|
+
e,
|
|
1273
|
+
s.transStart,
|
|
1274
|
+
r.toVtime,
|
|
1275
|
+
u.length
|
|
1276
|
+
);
|
|
1277
|
+
return { diffPath: u[p] ?? u[u.length - 1], label: `${r.from}→${r.to}` };
|
|
1278
|
+
}
|
|
1279
|
+
return { diffPath: Oe(r.to), label: r.to };
|
|
1280
|
+
}
|
|
1281
|
+
}
|
|
1282
|
+
const i = Jt(t, e);
|
|
1283
|
+
return i === "sil" ? { diffPath: null, label: "sil" } : { diffPath: Oe(i), label: i };
|
|
1284
|
+
}
|
|
1285
|
+
function ut(t, e) {
|
|
1286
|
+
const n = e ?? Ze(t.expressionPath);
|
|
1287
|
+
return `${Ze(t.mouthPath)}|${n}`;
|
|
1288
|
+
}
|
|
1289
|
+
function Nn(t, e, n, r, i, a, s, c) {
|
|
1290
|
+
const u = Hn(e, n), p = {
|
|
1291
|
+
mouthPath: u.diffPath,
|
|
1292
|
+
expressionPath: r
|
|
1293
|
+
}, _ = ut(p, i);
|
|
1294
|
+
_ !== a.lastDrawnKey && (a.lastDrawnKey = _, c(u.label), s(t, p));
|
|
1295
|
+
}
|
|
1296
|
+
function Wn(t, e, n, r, i, a, s = null) {
|
|
1297
|
+
const c = { mouthPath: null, expressionPath: e }, u = s != null ? `|idle@${Math.round(s * 1e3)}` : "", p = ut(c, n) + u;
|
|
1298
|
+
!(s != null) && p === r.lastDrawnKey || (r.lastDrawnKey = p, a(
|
|
1299
|
+
s != null ? "idle" : "sil"
|
|
1300
|
+
), i(t, c));
|
|
1301
|
+
}
|
|
1302
|
+
function Kn(t) {
|
|
1303
|
+
return new Worker(
|
|
1304
|
+
"/assets/visemeDiffPreview.worker-B8Juk7ys.js",
|
|
1305
|
+
{
|
|
1306
|
+
name: t?.name
|
|
1307
|
+
}
|
|
1308
|
+
);
|
|
1309
|
+
}
|
|
1310
|
+
function qn() {
|
|
1311
|
+
return typeof Worker < "u" && typeof OffscreenCanvas < "u";
|
|
1312
|
+
}
|
|
1313
|
+
function jn(t, e) {
|
|
1314
|
+
(t.width !== e.width || t.height !== e.height) && (t.width = e.width, t.height = e.height);
|
|
1315
|
+
const n = t.getContext("2d");
|
|
1316
|
+
n && n.drawImage(e, 0, 0);
|
|
1317
|
+
}
|
|
1318
|
+
function et(t, e) {
|
|
1319
|
+
let n = 0;
|
|
1320
|
+
return {
|
|
1321
|
+
usesWorker: !1,
|
|
1322
|
+
async loadAssets(r) {
|
|
1323
|
+
await be(r);
|
|
1324
|
+
},
|
|
1325
|
+
async renderSilOnly() {
|
|
1326
|
+
n += 1, await Fn(t);
|
|
1327
|
+
},
|
|
1328
|
+
async drawLiveFrame(r, i) {
|
|
1329
|
+
await be(), await Le(r, i);
|
|
1330
|
+
},
|
|
1331
|
+
async renderViseme(r, i) {
|
|
1332
|
+
n += 1;
|
|
1333
|
+
const a = n;
|
|
1334
|
+
e.onStatus?.(i), await Le(r, {
|
|
1335
|
+
mouthPath: Oe(i),
|
|
1336
|
+
expressionPath: null
|
|
1337
|
+
});
|
|
1338
|
+
},
|
|
1339
|
+
dispose() {
|
|
1340
|
+
n += 1;
|
|
1341
|
+
}
|
|
1342
|
+
};
|
|
1343
|
+
}
|
|
1344
|
+
function zn(t, e) {
|
|
1345
|
+
const n = new Kn();
|
|
1346
|
+
let r = 1, i = 0, a = !1, s = !1;
|
|
1347
|
+
const c = /* @__PURE__ */ new Map(), u = (o, f) => {
|
|
1348
|
+
s || n.postMessage(o, []);
|
|
1349
|
+
}, p = (o, f, d = !1) => {
|
|
1350
|
+
if (s) return Promise.resolve();
|
|
1351
|
+
const k = r++;
|
|
1352
|
+
return new Promise((x, F) => {
|
|
1353
|
+
c.set(k, {
|
|
1354
|
+
resolve: () => x(),
|
|
1355
|
+
reject: F,
|
|
1356
|
+
generation: typeof o.generation == "number" ? o.generation : void 0,
|
|
1357
|
+
expectRenderDone: d
|
|
1358
|
+
}), u({ ...o, requestId: k });
|
|
1359
|
+
});
|
|
1360
|
+
};
|
|
1361
|
+
n.onmessage = (o) => {
|
|
1362
|
+
if (s) return;
|
|
1363
|
+
const f = o.data;
|
|
1364
|
+
if (f.type === "status") {
|
|
1365
|
+
e.onStatus?.(f.label);
|
|
1366
|
+
return;
|
|
1367
|
+
}
|
|
1368
|
+
if (f.type === "frame") {
|
|
1369
|
+
const k = f.generation < 0;
|
|
1370
|
+
if (!k && f.generation !== i) {
|
|
1371
|
+
f.bitmap.close();
|
|
1372
|
+
return;
|
|
1373
|
+
}
|
|
1374
|
+
if (k && e.shouldAcceptLiveFrame && !e.shouldAcceptLiveFrame()) {
|
|
1375
|
+
f.bitmap.close();
|
|
1376
|
+
return;
|
|
1377
|
+
}
|
|
1378
|
+
jn(t, f.bitmap), f.bitmap.close();
|
|
1379
|
+
return;
|
|
1380
|
+
}
|
|
1381
|
+
const d = c.get(f.requestId);
|
|
1382
|
+
if (d)
|
|
1383
|
+
switch (f.type) {
|
|
1384
|
+
case "ready":
|
|
1385
|
+
case "loadAssetsDone":
|
|
1386
|
+
c.delete(f.requestId), d.resolve();
|
|
1387
|
+
break;
|
|
1388
|
+
case "renderDone":
|
|
1389
|
+
if (d.expectRenderDone && d.generation != null && f.generation !== d.generation)
|
|
1390
|
+
return;
|
|
1391
|
+
c.delete(f.requestId), d.resolve();
|
|
1392
|
+
break;
|
|
1393
|
+
case "renderAborted":
|
|
1394
|
+
c.delete(f.requestId), d.resolve();
|
|
1395
|
+
break;
|
|
1396
|
+
case "error":
|
|
1397
|
+
c.delete(f.requestId), e.onError?.(f.message), d.reject(new Error(f.message));
|
|
1398
|
+
break;
|
|
1399
|
+
}
|
|
1400
|
+
}, n.onerror = (o) => {
|
|
1401
|
+
if (s) return;
|
|
1402
|
+
const f = o.message || "Worker error";
|
|
1403
|
+
e.onError?.(f);
|
|
1404
|
+
for (const [, d] of c)
|
|
1405
|
+
d.reject(new Error(f));
|
|
1406
|
+
c.clear();
|
|
1407
|
+
};
|
|
1408
|
+
const _ = (async () => {
|
|
1409
|
+
await p({ type: "init" }), a = !0;
|
|
1410
|
+
})();
|
|
1411
|
+
return {
|
|
1412
|
+
usesWorker: !0,
|
|
1413
|
+
async loadAssets(o) {
|
|
1414
|
+
if (!s && (await _, !s)) {
|
|
1415
|
+
if (!a) throw new Error("Worker init failed");
|
|
1416
|
+
await p({
|
|
1417
|
+
type: "loadAssets",
|
|
1418
|
+
keyHex: o,
|
|
1419
|
+
urls: re()
|
|
1420
|
+
});
|
|
1421
|
+
}
|
|
1422
|
+
},
|
|
1423
|
+
async renderSilOnly() {
|
|
1424
|
+
if (s || (await _, s)) return;
|
|
1425
|
+
i += 1;
|
|
1426
|
+
const o = i;
|
|
1427
|
+
await p(
|
|
1428
|
+
{ type: "renderSil", generation: o },
|
|
1429
|
+
void 0,
|
|
1430
|
+
!0
|
|
1431
|
+
);
|
|
1432
|
+
},
|
|
1433
|
+
async drawLiveFrame(o, f) {
|
|
1434
|
+
s || (await _, !s && await p(
|
|
1435
|
+
{
|
|
1436
|
+
type: "drawFrame",
|
|
1437
|
+
diffPath: f.mouthPath,
|
|
1438
|
+
expressionDiffPath: f.expressionPath
|
|
1439
|
+
},
|
|
1440
|
+
void 0,
|
|
1441
|
+
!0
|
|
1442
|
+
));
|
|
1443
|
+
},
|
|
1444
|
+
async renderViseme(o, f, d = {}) {
|
|
1445
|
+
if (s || (await _, s)) return;
|
|
1446
|
+
i += 1;
|
|
1447
|
+
const k = i;
|
|
1448
|
+
try {
|
|
1449
|
+
await p(
|
|
1450
|
+
{
|
|
1451
|
+
type: "renderViseme",
|
|
1452
|
+
generation: k,
|
|
1453
|
+
to: f,
|
|
1454
|
+
from: d.from,
|
|
1455
|
+
transitionDurationMs: d.transitionDurationMs,
|
|
1456
|
+
gapMs: d.gapMs,
|
|
1457
|
+
threshold: d.threshold
|
|
1458
|
+
},
|
|
1459
|
+
void 0,
|
|
1460
|
+
!0
|
|
1461
|
+
);
|
|
1462
|
+
} catch (x) {
|
|
1463
|
+
if (x?.name === "AbortError") return;
|
|
1464
|
+
throw x;
|
|
1465
|
+
}
|
|
1466
|
+
},
|
|
1467
|
+
dispose() {
|
|
1468
|
+
if (!s) {
|
|
1469
|
+
s = !0, i += 1;
|
|
1470
|
+
for (const [, o] of c)
|
|
1471
|
+
o.resolve();
|
|
1472
|
+
c.clear(), n.terminate();
|
|
1473
|
+
}
|
|
1474
|
+
}
|
|
1475
|
+
};
|
|
1476
|
+
}
|
|
1477
|
+
function Gn(t, e = {}) {
|
|
1478
|
+
if (!qn())
|
|
1479
|
+
return et(t, e);
|
|
1480
|
+
try {
|
|
1481
|
+
return zn(t, e);
|
|
1482
|
+
} catch {
|
|
1483
|
+
return et(t, e);
|
|
1484
|
+
}
|
|
1485
|
+
}
|
|
1486
|
+
const Qn = {
|
|
1487
|
+
position: "relative",
|
|
1488
|
+
display: "inline-block",
|
|
1489
|
+
lineHeight: 0
|
|
1490
|
+
}, Jn = {
|
|
1491
|
+
display: "block",
|
|
1492
|
+
maxWidth: "100%",
|
|
1493
|
+
height: "auto"
|
|
1494
|
+
}, Xn = {
|
|
1495
|
+
position: "absolute",
|
|
1496
|
+
inset: 0,
|
|
1497
|
+
display: "flex",
|
|
1498
|
+
alignItems: "center",
|
|
1499
|
+
justifyContent: "center",
|
|
1500
|
+
padding: "1rem",
|
|
1501
|
+
background: "rgba(0,0,0,0.55)",
|
|
1502
|
+
color: "#fff",
|
|
1503
|
+
fontSize: "0.875rem",
|
|
1504
|
+
textAlign: "center"
|
|
1505
|
+
};
|
|
1506
|
+
function Ie(t) {
|
|
1507
|
+
return t === "cartesia" ? en : rt;
|
|
1508
|
+
}
|
|
1509
|
+
const tr = dt(function({
|
|
1510
|
+
id: e,
|
|
1511
|
+
assets: n,
|
|
1512
|
+
authToken: r,
|
|
1513
|
+
apiBase: i = $e,
|
|
1514
|
+
facesCdnBase: a = tt,
|
|
1515
|
+
tts: s = "google",
|
|
1516
|
+
voiceId: c,
|
|
1517
|
+
speakingRate: u = it,
|
|
1518
|
+
className: p,
|
|
1519
|
+
style: _,
|
|
1520
|
+
canvasStyle: o,
|
|
1521
|
+
onStatusChange: f,
|
|
1522
|
+
onDisplayStatus: d,
|
|
1523
|
+
onError: k,
|
|
1524
|
+
onReady: x,
|
|
1525
|
+
showErrorOverlay: F = !0
|
|
1526
|
+
}, Z) {
|
|
1527
|
+
const H = M(null), h = M(null), g = M(Vn()), P = M(In()), O = M(0), C = M(0), K = M(null), ie = M(!1), D = M(
|
|
1528
|
+
null
|
|
1529
|
+
), N = M(!1), [se, fe] = _e("idle"), [oe, de] = _e(null), [z, ke] = _e(!1), [he, ae] = _e(!1), [Ee, Ne] = _e({ width: 842, height: 1264 }), m = M(r ?? null), R = M(null), A = M(null), T = M(!1), b = M(x), I = M(d), q = M(k);
|
|
1530
|
+
b.current = x, I.current = d, q.current = k;
|
|
1531
|
+
const B = ee(
|
|
1532
|
+
(l) => {
|
|
1533
|
+
const y = l?.tts ?? s ?? A.current?.tts ?? "google", E = l?.voiceId ?? c ?? A.current?.voiceId ?? Ie(y);
|
|
1534
|
+
return { tts: y, voiceId: E };
|
|
1535
|
+
},
|
|
1536
|
+
[s, c]
|
|
1537
|
+
), v = ee((l) => {
|
|
1538
|
+
de(l), q.current?.(l);
|
|
1539
|
+
}, []), G = ee(
|
|
1540
|
+
(l) => {
|
|
1541
|
+
fe(l), f?.(l);
|
|
1542
|
+
},
|
|
1543
|
+
[f]
|
|
1544
|
+
), S = M(tn(G, i)), V = ee(() => {
|
|
1545
|
+
K.current != null && (clearTimeout(K.current), K.current = null);
|
|
1546
|
+
}, []), $ = ee(async () => {
|
|
1547
|
+
const l = C.current, y = h.current;
|
|
1548
|
+
!y || !z || S.current.isSpeaking() || (we(g.current), !N.current && (await y.renderSilOnly(), l !== C.current || S.current.isSpeaking()));
|
|
1549
|
+
}, [z]), J = ee(() => {
|
|
1550
|
+
V();
|
|
1551
|
+
const l = C.current;
|
|
1552
|
+
$(), K.current = setTimeout(() => {
|
|
1553
|
+
K.current = null, l === C.current && $();
|
|
1554
|
+
}, $n);
|
|
1555
|
+
}, [V, $]), ce = ee(async () => {
|
|
1556
|
+
const l = Math.floor(Date.now() / 1e3);
|
|
1557
|
+
if (R.current && l < R.current.expiresAt - 60)
|
|
1558
|
+
return R.current.key;
|
|
1559
|
+
const y = m.current ?? await Ke();
|
|
1560
|
+
m.current = y, S.current.setDeveloperToken(y);
|
|
1561
|
+
const E = await St(y, i);
|
|
1562
|
+
return R.current = {
|
|
1563
|
+
key: E.value,
|
|
1564
|
+
expiresAt: E.expiresAt
|
|
1565
|
+
}, E.value;
|
|
1566
|
+
}, [i]);
|
|
1567
|
+
le(() => {
|
|
1568
|
+
r && (m.current = r, S.current.setDeveloperToken(r));
|
|
1569
|
+
}, [r]), le(() => {
|
|
1570
|
+
S.current.setTtsProvider(s), A.current && (A.current.tts = s);
|
|
1571
|
+
}, [s]), le(() => {
|
|
1572
|
+
const l = c ?? Ie(s);
|
|
1573
|
+
A.current && (A.current.voiceId = l);
|
|
1574
|
+
}, [c, s]);
|
|
1575
|
+
const U = ht(
|
|
1576
|
+
() => n ? JSON.stringify(n) : `id:${e ?? ""}:${a}`,
|
|
1577
|
+
[n, e, a]
|
|
1578
|
+
);
|
|
1579
|
+
return le(() => {
|
|
1580
|
+
if (!e && !n) {
|
|
1581
|
+
v("AiTwin requires either `id` or `assets`");
|
|
1582
|
+
return;
|
|
1583
|
+
}
|
|
1584
|
+
let l = !1;
|
|
1585
|
+
T.current = !1, ke(!1), ae(!1), de(null), A.current = null, D.current = null, N.current = !1;
|
|
1586
|
+
const y = H.current;
|
|
1587
|
+
if (!y) return;
|
|
1588
|
+
const E = Gn(y, {
|
|
1589
|
+
onStatus: (w) => {
|
|
1590
|
+
l || I.current?.(w);
|
|
1591
|
+
},
|
|
1592
|
+
onError: (w) => {
|
|
1593
|
+
l || v(w);
|
|
1594
|
+
},
|
|
1595
|
+
shouldAcceptLiveFrame: () => S.current.isSpeaking()
|
|
1596
|
+
});
|
|
1597
|
+
return h.current = E, (async () => {
|
|
1598
|
+
try {
|
|
1599
|
+
let w, L, Q;
|
|
1600
|
+
if (n)
|
|
1601
|
+
w = n, L = s, Q = c ?? Ie(L);
|
|
1602
|
+
else {
|
|
1603
|
+
const Y = await wt(e);
|
|
1604
|
+
if (l) return;
|
|
1605
|
+
L = Y.tts, Q = c ?? Y.voiceId ?? Ie(L), w = It(Y.faceId, a);
|
|
1606
|
+
}
|
|
1607
|
+
A.current = { tts: L, voiceId: Q }, Rt(w), D.current = Rn(Dt());
|
|
1608
|
+
const X = m.current ?? r ?? await Ke();
|
|
1609
|
+
if (l) return;
|
|
1610
|
+
m.current = X, S.current.setDeveloperToken(X), S.current.setTtsProvider(L);
|
|
1611
|
+
const Se = ne() ? await ce() : void 0;
|
|
1612
|
+
if (l) return;
|
|
1613
|
+
const j = await hn(Se);
|
|
1614
|
+
if (!l && j.width > 0 && j.height > 0 && Ne(j), await E.loadAssets(Se), l) return;
|
|
1615
|
+
let ue = !1;
|
|
1616
|
+
try {
|
|
1617
|
+
await D.current.load(), ue = !0, N.current = !0, ae(!0);
|
|
1618
|
+
} catch (Y) {
|
|
1619
|
+
N.current = !1, ae(!1), console.warn("[AiTwin] Idle video unavailable:", Y);
|
|
1620
|
+
}
|
|
1621
|
+
if (!ue)
|
|
1622
|
+
await E.renderSilOnly(), I.current?.("sil");
|
|
1623
|
+
else {
|
|
1624
|
+
we(g.current);
|
|
1625
|
+
const Y = D.current.getVideo();
|
|
1626
|
+
D.current.setActive(!0);
|
|
1627
|
+
try {
|
|
1628
|
+
await Ye(y, Y, null), I.current?.("idle");
|
|
1629
|
+
} catch (me) {
|
|
1630
|
+
console.warn("[AiTwin] Initial idle frame paint failed:", me);
|
|
1631
|
+
}
|
|
1632
|
+
}
|
|
1633
|
+
if (l) return;
|
|
1634
|
+
T.current = !0, ke(!0), b.current?.();
|
|
1635
|
+
} catch (w) {
|
|
1636
|
+
if (l) return;
|
|
1637
|
+
if (w instanceof Pe) {
|
|
1638
|
+
v(w.message);
|
|
1639
|
+
return;
|
|
1640
|
+
}
|
|
1641
|
+
v(w instanceof Error ? w.message : "Failed to load AI twin");
|
|
1642
|
+
}
|
|
1643
|
+
})(), () => {
|
|
1644
|
+
l = !0, E.dispose(), h.current = null;
|
|
1645
|
+
};
|
|
1646
|
+
}, [U, e, a, r, i, ce, v]), le(() => {
|
|
1647
|
+
se === "done" && z && !S.current.isSpeaking() && J();
|
|
1648
|
+
}, [se, z, J]), le(() => {
|
|
1649
|
+
const l = S.current, y = { current: !1 }, E = () => {
|
|
1650
|
+
const w = H.current, L = h.current;
|
|
1651
|
+
if (!w || !L || !z) {
|
|
1652
|
+
O.current = requestAnimationFrame(E);
|
|
1653
|
+
return;
|
|
1654
|
+
}
|
|
1655
|
+
const Q = l.isSpeaking(), X = l.getPlaybackElapsedMs(), Se = Q ? xn(
|
|
1656
|
+
X,
|
|
1657
|
+
l.getSentenceEmotions(),
|
|
1658
|
+
l.getWordQueue(),
|
|
1659
|
+
l.getStreamMood()
|
|
1660
|
+
) : l.getStreamMood();
|
|
1661
|
+
if (Q) {
|
|
1662
|
+
if (N.current && D.current?.setActive(!1), l.getVisemeQueue().length === 0) {
|
|
1663
|
+
O.current = requestAnimationFrame(E);
|
|
1664
|
+
return;
|
|
1665
|
+
}
|
|
1666
|
+
P.current.setTargetEmotion(Se), P.current.advance(performance.now());
|
|
1667
|
+
const ue = P.current.getExpressionPath(), Y = P.current.getDrawKey();
|
|
1668
|
+
Nn(
|
|
1669
|
+
w,
|
|
1670
|
+
l.getVisemeQueue(),
|
|
1671
|
+
X,
|
|
1672
|
+
ue,
|
|
1673
|
+
Y,
|
|
1674
|
+
g.current,
|
|
1675
|
+
(me, lt) => L.drawLiveFrame(me, lt),
|
|
1676
|
+
(me) => I.current?.(me)
|
|
1677
|
+
);
|
|
1678
|
+
} else if (N.current && D.current) {
|
|
1679
|
+
const j = D.current;
|
|
1680
|
+
j.setActive(!0), y.current || (y.current = !0, Ye(w, j.getVideo()).then(() => I.current?.("idle")).finally(() => {
|
|
1681
|
+
y.current = !1;
|
|
1682
|
+
}));
|
|
1683
|
+
} else
|
|
1684
|
+
Wn(
|
|
1685
|
+
w,
|
|
1686
|
+
null,
|
|
1687
|
+
"__sil__",
|
|
1688
|
+
g.current,
|
|
1689
|
+
(j, ue) => L.drawLiveFrame(j, ue),
|
|
1690
|
+
(j) => I.current?.(j)
|
|
1691
|
+
);
|
|
1692
|
+
ie.current && !Q && (P.current.reset(), N.current && D.current?.restart(), J()), ie.current = Q, O.current = requestAnimationFrame(E);
|
|
1693
|
+
};
|
|
1694
|
+
return O.current = requestAnimationFrame(E), () => {
|
|
1695
|
+
cancelAnimationFrame(O.current), V(), l.stop();
|
|
1696
|
+
};
|
|
1697
|
+
}, [z, he, J, V]), mt(
|
|
1698
|
+
Z,
|
|
1699
|
+
() => ({
|
|
1700
|
+
speakText: async (l, y) => {
|
|
1701
|
+
const E = l.trim();
|
|
1702
|
+
if (!E) return;
|
|
1703
|
+
const w = h.current;
|
|
1704
|
+
if (!w || !T.current && !z)
|
|
1705
|
+
throw new Error("AI twin is not ready");
|
|
1706
|
+
const { tts: L, voiceId: Q } = B(y);
|
|
1707
|
+
de(null), C.current += 1, V(), we(g.current), P.current.reset(), ie.current = !1;
|
|
1708
|
+
try {
|
|
1709
|
+
N.current ? (D.current?.setActive(!1), D.current?.restart(), we(g.current)) : await w.renderSilOnly(), await S.current.speak(E, {
|
|
1710
|
+
voiceId: Q,
|
|
1711
|
+
speakingRate: y?.speakingRate ?? u,
|
|
1712
|
+
tts: L
|
|
1713
|
+
});
|
|
1714
|
+
} catch (X) {
|
|
1715
|
+
throw v(X instanceof Error ? X.message : "TTS failed"), X;
|
|
1716
|
+
}
|
|
1717
|
+
},
|
|
1718
|
+
stop: () => {
|
|
1719
|
+
C.current += 1, V(), S.current.stop(), we(g.current), ie.current = !1, J();
|
|
1720
|
+
},
|
|
1721
|
+
setTtsProvider: (l) => {
|
|
1722
|
+
S.current.setTtsProvider(l), A.current && (A.current.tts = l);
|
|
1723
|
+
},
|
|
1724
|
+
renderViseme: async (l, y) => {
|
|
1725
|
+
const E = h.current, w = H.current;
|
|
1726
|
+
if (!E || !w || !T.current)
|
|
1727
|
+
throw new Error("AI twin is not ready");
|
|
1728
|
+
await E.renderViseme(w, l, {
|
|
1729
|
+
from: y?.from,
|
|
1730
|
+
transitionDurationMs: y?.transitionDurationMs,
|
|
1731
|
+
gapMs: y?.gapMs,
|
|
1732
|
+
onStatus: (L) => I.current?.(L)
|
|
1733
|
+
});
|
|
1734
|
+
},
|
|
1735
|
+
isReady: () => T.current,
|
|
1736
|
+
getStatus: () => se
|
|
1737
|
+
}),
|
|
1738
|
+
[
|
|
1739
|
+
z,
|
|
1740
|
+
u,
|
|
1741
|
+
V,
|
|
1742
|
+
J,
|
|
1743
|
+
v,
|
|
1744
|
+
B,
|
|
1745
|
+
se
|
|
1746
|
+
]
|
|
1747
|
+
), /* @__PURE__ */ ft("div", { className: p, style: { ...Qn, ..._ }, children: [
|
|
1748
|
+
/* @__PURE__ */ We(
|
|
1749
|
+
"canvas",
|
|
1750
|
+
{
|
|
1751
|
+
ref: H,
|
|
1752
|
+
width: Ee.width,
|
|
1753
|
+
height: Ee.height,
|
|
1754
|
+
style: { ...Jn, ...o },
|
|
1755
|
+
"aria-label": e ? `AI twin ${e}` : "AI twin face"
|
|
1756
|
+
}
|
|
1757
|
+
),
|
|
1758
|
+
F && oe ? /* @__PURE__ */ We("div", { style: Xn, children: oe }) : null
|
|
1759
|
+
] });
|
|
1760
|
+
});
|
|
1761
|
+
export {
|
|
1762
|
+
tr as AiTwin,
|
|
1763
|
+
Pe as AiTwinNotFoundError,
|
|
1764
|
+
$e as DEFAULT_API_BASE,
|
|
1765
|
+
tt as DEFAULT_FACES_CDN_BASE,
|
|
1766
|
+
Mt as OCULUS_VISEME_IDS,
|
|
1767
|
+
Zt as SPEAKING_RATE_MAX,
|
|
1768
|
+
Yt as SPEAKING_RATE_MIN,
|
|
1769
|
+
er as VISEME_IDS,
|
|
1770
|
+
en as VISEME_TEST_CARTESIA_VOICE_ID,
|
|
1771
|
+
it as VISEME_TEST_SPEAKING_RATE,
|
|
1772
|
+
rt as VISEME_TEST_VOICE_ID,
|
|
1773
|
+
tn as createAvatarTtsLipsyncController,
|
|
1774
|
+
wt as fetchAiTwin,
|
|
1775
|
+
Ke as fetchDevAuthToken,
|
|
1776
|
+
Ut as normalizeOculusViseme,
|
|
1777
|
+
Jt as resolveVisemeAtTime,
|
|
1778
|
+
Xt as resolveWordAtTime
|
|
1779
|
+
};
|